You are on page 1of 138

Occasional Paper 13 - DECENTRALISATION: A SURVEY OF LITERATURE FROM A HUMAN DEVELOPMENT PERSPECTIVE

DECENTRALISATION: A SURVEY OF LITERATURE FROM A HUMAN DEVELOPMENT PERSPECTIVE


Jeni Klugman, 1994

PART I Overview and Conclusions PART II 1. 2. 3. 4. 5. 6. 7. Introduction Efficiency Resource Availability Participation The Impact Upon Economic Activity Equity Obstacles

PART III - SECTORAL SPECIFIC CONSIDERATIONS 1. 2. 3. 4. 5. Education Health Water and sanitation Housing Economic infrastructure

PART IV- EXPENDITURE ALLOCATION - COMPARATIVE PRIORITIES 1. 2. 3. 4. Introduction and summary Local community priorities Local government priorities Some Empirical Evidence Bibliography

This paper was completed in late 1991, and may not reflect current national policies and practices. Helpful comments were received from Gustav Ranis and Frances Stewart. The views expressed here do not represent those of the World Bank nor the UNDP.

OVERVIEW AND CONCLUSIONS Decentralization has been a popular theme in development thinking and practice for at least two decades. The interest here lies in the impact of decentralisation of government expenditures and revenues upon human development. This survey explores the extensive literature on decentralisation, revealing the lack of quantitative and rigorous studies. It emerges that this is an area where few, if any, general propositions will hold. Nonetheless detailed analysis of the various dimensions of decentralisation - participation, financing and comparative priorities - and of the relevant effects upon efficiency, resource availability and equity, enables us to draw some interesting lessons from its theory and practice. These can be summarised as follows. Regarding efficiency, the main problems associated with centralisation include the irrelevance of projects and policies to local needs and conditions, lack of adaptability, delays, under-utilisation of local resources and inadequate maintenance. It is observed that decentralisation has demonstrated potential to lead to more appropriate and better utilised facilities, lower costs per unit of service and improved operations and maintenance. However, greater efficiency need not necessarily accompany decentralisation, given the risk of loss of economies of scale, duplication and overlap. In any case, there must be a fairly thoroughgoing devolution of authority in order that the presumed advantages might follow. We distinguish two lines of thought with respect to political participation in public decision making, which focus upon the developmental and instrumental aspects, i.e. as an end in itself, and as a means to better outcomes, respectively. Whilst acknowledging the value of the former, our ultimate interest here lies in the impact of decentralisation and participation upon the nature of expenditure and revenue decisions affecting human development. Even where national officials and local elites dominate local decision making structures, the outcomes may be more positive in terms of human development than that which would have resulted under a centralised system. More extensive evidence bearing upon these issues is discussed under the headings of comparative priorities and equity. The discussion and quantitative analysis of the comparative priorities of different levels of government suggests that local governments, where some thoroughgoing devolution of decision making power takes place, are likely to make expenditure decisions which accord to the basic priorities of human development. Not only do local governments tend to focus on more basic (and cheaper) levels of provision, the evidence suggests that their expenditure patterns as a whole are more oriented towards education, community

amenities and health care. This is contrasted to a picture of central priorities, wherein the social allocation ratio averages 28%, and the social priority ratio 38% (HDR 1991, p.41). With respect to effects on equity, a distinction is drawn between its inter- and intradistrict aspects. It is found that intra-district equity is likely to improve, following increased expenditure on basic social services which increases access and utilisation on the part of poorer members of the community. At the same time however, the pivotal role of central grants in affecting inter-district equity is identified and explored, in both theoretical and practical terms. It is found that the prevailing fiscal arrangements under decentralisation - by design or otherwise - typically fail to enhance inter-district equity. The incidence of taxation is also investigated. Due to data limitations, no conclusive determination can be made as to the relative progressiveness of decentralised, versus centralised taxation systems; the most which could be said that neither appears to be very regressive. Beyond the expansion of basic social services, increasing income earning opportunities is a crucial aspect of human development. In principle, decentralisation may promote economic activity via several routes including an increased infusion of capital and other resources, the more extensive provision of infrastructure, and a more effective enabling environment than would have been the case under a centralised system. A number of examples are found to illustrate these possibilities. Resource availability under a decentralised system is an important aspect of the present survey. It is shown that the nature of the assignment problem means that vertical fiscal imbalances are almost inherently likely to occur, and that in practice local taxation is limited by the fragmented nature of the system, centrally imposed restrictions and administrative constraints. At the same time it is clear that taxation at all levels is subject to evasion and practical administrative problems. It is argued that cost recovery raises significant efficiency and equity considerations, which lead to the conclusion - contrary to World Bank opinion - that the nature of local government responsibilities (e.g. primary education and primary health care) renders user charges inappropriate at the local level. Given the likelihood of fiscal gaps emerging, inter governmental financial arrangements will play an important role. These are explored, with particular interest being accorded to the principle of financial autonomy. Whilst considerations of equity point to the need for central grants, the risk of national government dominance and corresponding local vulnerability is very real, as illustrated by a number of case studies. The need for increased public expenditure on human development leads us to explore the possibility that decentralisation may increase total public resource mobilisation. It is found that voluntarism is fairly widespread at the local level (as predicted by the visible benefits principle) although it rarely accounts for substantial amounts of revenue. Borrowing is another route by which additional resources can be mobilised for local government, although this can be problematic given, for example, the potentially adverse implications for the efficacy of national macroeconomic policy. It is likely to remain a relatively marginal form of finance in developing countries, limited to the largest and most financially viable municipalities. It is shown that inter governmental transfers often

tend to provide disincentives to, and/or substitute for, local revenue raising especially in the absence of such mechanisms as matching requirements. However the empirical evidence available suggests that, in general, the degree of decentralisation and total public resource mobilisation have little to do with each other. Finally the significant underlying theme relating to the obstacles confronted in decentralisation efforts is explored. There is ample evidence of political resistance to decentralisation by those who perceive their vested interests to be threatened, and of the bureaucratic problems, especially relating to staffing. Overall the picture of decentralisation in developing countries which emerges reflects both in theory and practice - significant potential for human development. Provided that the devolution of decision making power is thoroughgoing, the provision and utilisation of basic social services and levels of economic activity can be expected to expand. This does not deny a significant role for national authorities - whose actions have a crucial bearing in ensuring a minimum degree of provision (in both qualitative and quantitative terms), acting to take advantage of economies of scale, and avoiding unnecessary overlap and duplication, and in mobilising public revenue, as well as promoting inter district equity. It may therefore be appropriate that the focus of central governments shift, from direct decision making about the details of provision of basic social and economic infrastructure to providing a supportive role where local institutions can more effectively provide such services, especially in previously neglected regions, and to poorer sections of the community. This is not to suggest that the contemporary experiences of decentralisation in developing countries typically present such a rosy picture. It is seen that under existing arrangements the functions and finances of local government institutions are often quite limited, and so more appropriately described as deconcentration; that local administrative and revenue raising capacity is inadequate; and that inter-district equity may well worsen. Thus the message portrayed here is that decentralisation, under certain significant conditions, can have substantial positive implications for human development in developing countries. 1. Introduction Human development, defined in terms of the expansion of peoples' capabilities, has several aspects. Foremost among these are to lead a long and healthy life, to acquire knowledge and to enjoy a decent standard of living. These, in turn, raise such dimensions as access to basic social services, expansion of income-earning opportunities; and participation - both political and economic. The task here is to evaluate the literature on decentralisation in the above terms. Decentralisation can be seen in the light of various objectives which, in turn, contribute to human development. The comparative priorities of different levels of government, together with issues of equity, efficiency and financing are all crucial. There are also important questions, very relevant to human development, about participation in decision

making. In Part II this survey reviews the evidence on these dimensions, in order to appraise the extent to which decentralisation can contribute to human development, in terms of the above-mentioned dimensions. Part III investigates the characteristics of, and the issues raised by decentralisation in specific sectors. Finally Part IV tackles the issue of comparative priorities at different levels of government. It becomes clear that few definitive propositions hold in this area - in the sense that the question is not simply whether decentralisation per se is 'good' or 'bad' for human development. Rather there should be an appreciation of the various possibilities which can emerge under the rubric of decentralisation - the questions are therefore how to decentralise and what to decentralise and to whom, in order to maximise the benefits for human development. It is also important to bear the comparative aspect in mind - it is of fundamental interest to learn how local governments perform, relative to central government. Defining decentralisation There are many kinds of reforms which fall under the banner of decentralisation. It is therefore useful to follow Hicks (1961) and Rondinelli (1984) and distinguish between different modes of decentralisation along the following lines:

deconcentration - spatial relocation of decision making - i.e. the transfer of some administrative responsibility or authority to lower levels within central government ministries or agencies; delegation - assignment of specific decision making authority - i.e. the transfer of managerial responsibility for specifically defined functions to public organisations (e.g. local governments or parastatals) outside the normal bureaucratic structure of central government; and devolution - transfer of responsibility for governing, understood more broadly i.e. the creation or strengthening, financially or legally, of sub-national units of governments, whose activities are substantially outside the direct control of central government.

As far as possible, this survey will focus upon the second and third modes of decentralisation -although as will be seen, the first more limited form appears to be the most common approach reflected by developing country experience. There is frequently a large gap between the rhetoric, and the reality of the extent to which national governments are prepared to decentralise decision making authority. (Conyers (1981,1986) emphasises this point, as does Slater (1989).) Moreover, it can be argued that many decentralisation programs have involved devolution to "hybrid" bodies, composed of both locally elected and centrally appointed representatives (Conyers, 1981). It is possible to find so-called devolved "local governments" that have similar, or even less discretionary authority than do deconcentrated units" (Olowu 1989). And it is difficult to draw a sharp line between devolution and deconcentration when, for example,

executive officials are popularly elected (as are town and district officers in Tonga) (Larmour 1983). It is therefore important to examine attempts at decentralisation carefully. The Extent of Decentralisation The impact of decentralisation upon human development will obviously depend upon the relative importance of local government. In the present context, this is best indicated by the share of local government in total government expenditure, which varies widely among developing countries. This measure is only a rough guide, however, in the sense that local spending decisions may not be autonomous, as discussed below. One survey, for a sample of sixteen developing countries for which comparable data was available, found that the share ranged from 55% in India, down to 2.5% in The Gambia (WDR 1988). For seven of those countries, less that 10% of total government spending was conducted through local government. The size of the sub-national units is a related dimension. The main focus of the present survey is upon what is typically the lowest level of government - often known as the district or area. This can vary quite widely in size, even within countries. In Botswana, the range in population is from 22 000 to 364 000. In Africa generally, the size (both in terms of area and population) of the basic units of government is relatively larger than that of industrial countries (Olowu 1989) (see Table 1). Obviously the demographic and geographic circumstances of the particular country are, amongst other factors, relevant as extreme alternatives compare Indonesia or India to, say, Costa Rica or The Gambia. There may also be some relationship between, for example, the relative efficiency of local government performance and its size, due to such factors as economies of scale. Throughout this survey there are also references to what is often the second tier of government - state or provincial. This is due to (i) the need for evidence on particular aspects of decentralisation (e.g. the assignment problem); and (ii) the implications of such arrangements often appear to be relevant to decentralisation to lower levels of government.
Occasional Paper 13 - DECENTRALISATION: A SURVEY OF LITERATURE FROM A HUMAN DEVELOPMENT PERSPECTIVE

DECENTRALISATION: A SURVEY OF LITERATURE FROM A HUMAN DEVELOPMENT PERSPECTIVE


Jeni Klugman, 1994

PART I Overview and Conclusions PART II 1. 2. 3. 4. 5. 6. 7. Introduction Efficiency Resource Availability Participation The Impact Upon Economic Activity Equity Obstacles

PART III - SECTORAL SPECIFIC CONSIDERATIONS 1. 2. 3. 4. 5. Education Health Water and sanitation Housing Economic infrastructure

PART IV- EXPENDITURE ALLOCATION - COMPARATIVE PRIORITIES 1. 2. 3. 4. Introduction and summary Local community priorities Local government priorities Some Empirical Evidence Bibliography

This paper was completed in late 1991, and may not reflect current national policies and practices. Helpful comments were received from Gustav Ranis and Frances Stewart. The views expressed here do not represent those of the World Bank nor the UNDP.

OVERVIEW AND CONCLUSIONS Decentralization has been a popular theme in development thinking and practice for at least two decades. The interest here lies in the impact of decentralisation of government expenditures and revenues upon human development. This survey explores the extensive literature on decentralisation, revealing the lack of quantitative and rigorous studies. It emerges that this is an area where few, if any, general propositions will hold. Nonetheless

detailed analysis of the various dimensions of decentralisation - participation, financing and comparative priorities - and of the relevant effects upon efficiency, resource availability and equity, enables us to draw some interesting lessons from its theory and practice. These can be summarised as follows. Regarding efficiency, the main problems associated with centralisation include the irrelevance of projects and policies to local needs and conditions, lack of adaptability, delays, under-utilisation of local resources and inadequate maintenance. It is observed that decentralisation has demonstrated potential to lead to more appropriate and better utilised facilities, lower costs per unit of service and improved operations and maintenance. However, greater efficiency need not necessarily accompany decentralisation, given the risk of loss of economies of scale, duplication and overlap. In any case, there must be a fairly thoroughgoing devolution of authority in order that the presumed advantages might follow. We distinguish two lines of thought with respect to political participation in public decision making, which focus upon the developmental and instrumental aspects, i.e. as an end in itself, and as a means to better outcomes, respectively. Whilst acknowledging the value of the former, our ultimate interest here lies in the impact of decentralisation and participation upon the nature of expenditure and revenue decisions affecting human development. Even where national officials and local elites dominate local decision making structures, the outcomes may be more positive in terms of human development than that which would have resulted under a centralised system. More extensive evidence bearing upon these issues is discussed under the headings of comparative priorities and equity. The discussion and quantitative analysis of the comparative priorities of different levels of government suggests that local governments, where some thoroughgoing devolution of decision making power takes place, are likely to make expenditure decisions which accord to the basic priorities of human development. Not only do local governments tend to focus on more basic (and cheaper) levels of provision, the evidence suggests that their expenditure patterns as a whole are more oriented towards education, community amenities and health care. This is contrasted to a picture of central priorities, wherein the social allocation ratio averages 28%, and the social priority ratio 38% (HDR 1991, p.41). With respect to effects on equity, a distinction is drawn between its inter- and intradistrict aspects. It is found that intra-district equity is likely to improve, following increased expenditure on basic social services which increases access and utilisation on the part of poorer members of the community. At the same time however, the pivotal role of central grants in affecting inter-district equity is identified and explored, in both theoretical and practical terms. It is found that the prevailing fiscal arrangements under decentralisation - by design or otherwise - typically fail to enhance inter-district equity. The incidence of taxation is also investigated. Due to data limitations, no conclusive determination can be made as to the relative progressiveness of decentralised, versus centralised taxation systems; the most which could be said that neither appears to be very regressive.

Beyond the expansion of basic social services, increasing income earning opportunities is a crucial aspect of human development. In principle, decentralisation may promote economic activity via several routes including an increased infusion of capital and other resources, the more extensive provision of infrastructure, and a more effective enabling environment than would have been the case under a centralised system. A number of examples are found to illustrate these possibilities. Resource availability under a decentralised system is an important aspect of the present survey. It is shown that the nature of the assignment problem means that vertical fiscal imbalances are almost inherently likely to occur, and that in practice local taxation is limited by the fragmented nature of the system, centrally imposed restrictions and administrative constraints. At the same time it is clear that taxation at all levels is subject to evasion and practical administrative problems. It is argued that cost recovery raises significant efficiency and equity considerations, which lead to the conclusion - contrary to World Bank opinion - that the nature of local government responsibilities (e.g. primary education and primary health care) renders user charges inappropriate at the local level. Given the likelihood of fiscal gaps emerging, inter governmental financial arrangements will play an important role. These are explored, with particular interest being accorded to the principle of financial autonomy. Whilst considerations of equity point to the need for central grants, the risk of national government dominance and corresponding local vulnerability is very real, as illustrated by a number of case studies. The need for increased public expenditure on human development leads us to explore the possibility that decentralisation may increase total public resource mobilisation. It is found that voluntarism is fairly widespread at the local level (as predicted by the visible benefits principle) although it rarely accounts for substantial amounts of revenue. Borrowing is another route by which additional resources can be mobilised for local government, although this can be problematic given, for example, the potentially adverse implications for the efficacy of national macroeconomic policy. It is likely to remain a relatively marginal form of finance in developing countries, limited to the largest and most financially viable municipalities. It is shown that inter governmental transfers often tend to provide disincentives to, and/or substitute for, local revenue raising especially in the absence of such mechanisms as matching requirements. However the empirical evidence available suggests that, in general, the degree of decentralisation and total public resource mobilisation have little to do with each other. Finally the significant underlying theme relating to the obstacles confronted in decentralisation efforts is explored. There is ample evidence of political resistance to decentralisation by those who perceive their vested interests to be threatened, and of the bureaucratic problems, especially relating to staffing. Overall the picture of decentralisation in developing countries which emerges reflects both in theory and practice - significant potential for human development. Provided that the devolution of decision making power is thoroughgoing, the provision and utilisation of basic social services and levels of economic activity can be expected to expand. This does not deny a significant role for national authorities - whose actions have a crucial

bearing in ensuring a minimum degree of provision (in both qualitative and quantitative terms), acting to take advantage of economies of scale, and avoiding unnecessary overlap and duplication, and in mobilising public revenue, as well as promoting inter district equity. It may therefore be appropriate that the focus of central governments shift, from direct decision making about the details of provision of basic social and economic infrastructure to providing a supportive role where local institutions can more effectively provide such services, especially in previously neglected regions, and to poorer sections of the community. This is not to suggest that the contemporary experiences of decentralisation in developing countries typically present such a rosy picture. It is seen that under existing arrangements the functions and finances of local government institutions are often quite limited, and so more appropriately described as deconcentration; that local administrative and revenue raising capacity is inadequate; and that inter-district equity may well worsen. Thus the message portrayed here is that decentralisation, under certain significant conditions, can have substantial positive implications for human development in developing countries. 1. Introduction Human development, defined in terms of the expansion of peoples' capabilities, has several aspects. Foremost among these are to lead a long and healthy life, to acquire knowledge and to enjoy a decent standard of living. These, in turn, raise such dimensions as access to basic social services, expansion of income-earning opportunities; and participation - both political and economic. The task here is to evaluate the literature on decentralisation in the above terms. Decentralisation can be seen in the light of various objectives which, in turn, contribute to human development. The comparative priorities of different levels of government, together with issues of equity, efficiency and financing are all crucial. There are also important questions, very relevant to human development, about participation in decision making. In Part II this survey reviews the evidence on these dimensions, in order to appraise the extent to which decentralisation can contribute to human development, in terms of the above-mentioned dimensions. Part III investigates the characteristics of, and the issues raised by decentralisation in specific sectors. Finally Part IV tackles the issue of comparative priorities at different levels of government. It becomes clear that few definitive propositions hold in this area - in the sense that the question is not simply whether decentralisation per se is 'good' or 'bad' for human development. Rather there should be an appreciation of the various possibilities which can emerge under the rubric of decentralisation - the questions are therefore how to decentralise and what to decentralise and to whom, in order to maximise the benefits for human development. It is also important to bear the comparative aspect in mind - it is of fundamental interest to learn how local governments perform, relative to central government.

Defining decentralisation There are many kinds of reforms which fall under the banner of decentralisation. It is therefore useful to follow Hicks (1961) and Rondinelli (1984) and distinguish between different modes of decentralisation along the following lines:

deconcentration - spatial relocation of decision making - i.e. the transfer of some administrative responsibility or authority to lower levels within central government ministries or agencies; delegation - assignment of specific decision making authority - i.e. the transfer of managerial responsibility for specifically defined functions to public organisations (e.g. local governments or parastatals) outside the normal bureaucratic structure of central government; and devolution - transfer of responsibility for governing, understood more broadly i.e. the creation or strengthening, financially or legally, of sub-national units of governments, whose activities are substantially outside the direct control of central government.

As far as possible, this survey will focus upon the second and third modes of decentralisation -although as will be seen, the first more limited form appears to be the most common approach reflected by developing country experience. There is frequently a large gap between the rhetoric, and the reality of the extent to which national governments are prepared to decentralise decision making authority. (Conyers (1981,1986) emphasises this point, as does Slater (1989).) Moreover, it can be argued that many decentralisation programs have involved devolution to "hybrid" bodies, composed of both locally elected and centrally appointed representatives (Conyers, 1981). It is possible to find so-called devolved "local governments" that have similar, or even less discretionary authority than do deconcentrated units" (Olowu 1989). And it is difficult to draw a sharp line between devolution and deconcentration when, for example, executive officials are popularly elected (as are town and district officers in Tonga) (Larmour 1983). It is therefore important to examine attempts at decentralisation carefully. The Extent of Decentralisation The impact of decentralisation upon human development will obviously depend upon the relative importance of local government. In the present context, this is best indicated by the share of local government in total government expenditure, which varies widely among developing countries. This measure is only a rough guide, however, in the sense that local spending decisions may not be autonomous, as discussed below. One survey, for a sample of sixteen developing countries for which comparable data was available, found that the share ranged from 55% in India, down to 2.5% in The Gambia (WDR 1988). For seven of those countries, less that 10% of total government spending was conducted through local government.

The size of the sub-national units is a related dimension. The main focus of the present survey is upon what is typically the lowest level of government - often known as the district or area. This can vary quite widely in size, even within countries. In Botswana, the range in population is from 22 000 to 364 000. In Africa generally, the size (both in terms of area and population) of the basic units of government is relatively larger than that of industrial countries (Olowu 1989) (see Table 1). Obviously the demographic and geographic circumstances of the particular country are, amongst other factors, relevant as extreme alternatives compare Indonesia or India to, say, Costa Rica or The Gambia. There may also be some relationship between, for example, the relative efficiency of local government performance and its size, due to such factors as economies of scale. Throughout this survey there are also references to what is often the second tier of government - state or provincial. This is due to (i) the need for evidence on particular aspects of decentralisation (e.g. the assignment problem); and (ii) the implications of such arrangements often appear to be relevant to decentralisation to lower levels of government.
Occasional Paper 13 - DECENTRALISATION: A SURVEY OF LITERATURE FROM A HUMAN DEVELOPMENT PERSPECTIVE

2. Efficiency

Overview of the Theoretical Arguments Preferences Information Principals and Agents Cost Per Unit of Service Low Costs of Local Resources Costs of Scale Externalities Quality of Provision Implementation and Maintenance Local Participation in Decision Making Local Institutional Capacity The Comparative Extent of Malfeasance Efficiency - Sectoral Considerations Education Health Water and sanitation Housing Economic infrastructure

2. Efficiency

This section attempts to set out and appraise the impact of decentralisation upon efficiency. Efficiency is a multi-faceted concept which can be used in at least three senses. Allocative efficiency involves the consideration of what is produced, preferences and how it is allocated between agents. From a neo-classical economic point of view, the basic proposition is that local governments are better able, given differing local circumstances and preferences, to satisfy varying local demands for public goods and services. X- or technical efficiency refers to cost-minimisation. In practice attention tends to focus upon the relationships between decentralisation and unit costs, the use of local resources, technical adequacy or quality, and the comparative extent of malfeasance. It is also important to view efficiency over time, or the intertemporal allocation of resources, which brings the issues of implementation and maintenance to the fore. We proceed to examine the various dimensions of efficiency in turn, adducing specific evidence wherever possible. It is demonstrated that, as in many other aspects of decentralisation, efficiency raises countervailing factors. The design of public goods and services may be more in accordance with local preferences under a decentralised system (allocative efficiency), but weighing against this are central standards which can ensure a minimum degree of quality and quantity of provision. The utilisation of local resources, information and technology may lower costs (technical efficiency); but the existence of economies of scale points in favour of more centralised provision (provided that diseconomies do not thereby appear). The presence of externalities - i.e. absence of a market - also tends to justify provision and control by higher levels of government. Finally regarding implementation and maintenance (especially in the long term), the literature and evidence favour decentralisation, provided that this approach obtains a higher degree of user-participation than would have been the case in a centralised system. It follows that simple generalisations are not likely to emerge as to whether and how the structure of government affects efficiency. Nonetheless on balance it appears that theoretical and empirical considerations tend to favour local, rather than central responsibility. This is especially the case for public goods and services that have primarily local effects and where local adaptation is important. Overview of the Theoretical Arguments
"The State should offer those services that correspond to its citizens' preferences... decentralisation is better apt to take into account the different preferences of the community's members than are extremely unitary states with their systematically uniform approach." (Frey 1977)

Preferences The underlying economic theory falls into the realm of welfare economics and public choice (see Helm and Smith, 1987). It builds upon the propositions that individual preferences may vary by locality, and that information constraints inhibit the effective operation of central authorities. Thus the basic economic rationale for decentralisation rests largely on a negative observation, the inability of central government to deliver many public services efficiently. It has been suggested that the most efficient allocation

of public resources is attained if such services are provided (and paid for) by governments responsible to those most directly affected (Tiebout, 1956). This is based on following assumptions. First, that the demand for local public services varies from place to place, because tastes and willingness to pay differ, for geographic, cultural and historical reasons (and that preferences within each locality are reasonably homogeneous). If this is the case, the central provision of local public goods, (if it tends to be uniform across the country), is unlikely to please anybody. The second assumption is that decentralisation would result in every local government providing a different bundle of local public services, which reflects local preferences. Tiebout (1956) suggested that rather than attempt to provide a voting mechanism to allow individuals to express their preferences, such preferences might be reflected in where people choose to live - i.e. where the bundle of local public goods provided coincided with their preferred choices. This thesis obviously requires several strong assumptions about freedom of mobility and the nature of the migration decision, and its applicability is limited. "Voting with one's feet" is impeded considerably both by costs of migration which are higher than the advantages expected elsewhere, and by the "very fuzzy ideas" that individuals tend to have about the cost/benefit effects of their choice of residence (Buchanan 1974, cited in Frenkel 1986). Unfortunately, there is very little empirical research on the relationship between consumer satisfaction with public services and organisational arrangements. (The only evidence uncovered related to the U.S. context, where it was found that urban areas with small scale police organisations tended to satisfy residents more than did larger forces: Frenkel 1986). The argument about preferences is nonetheless useful in highlighting the possibility that local governments act more in accordance with the needs and priorities of local communities than would higher authorities. It gains empirical support from such areas as implementation and maintenance, where local involvement has been found to result in more appropriate and better utilised local public services. Levels of utilisation have a significant impact upon the cost per unit of service, particularly in the long term. In the water sector it has been found that in the absence of some thoroughgoing devolution of decision making power to users, the gap between local preferences and product design generally results in poor long term prospects for the services (see below). In the health sector, levels of utilisation are often low, partly due to difficulties associated with access (distance, time, expense etc.), and partly due to the lack of local receptivity to government provided services. Local conceptions of actions appropriate to promote good health and combat illness may well differ from those offered by public health officials. It is important that the design of health services takes this into account, so that the organisational form evokes "active and appropriate use and co-operation based upon comprehending acceptance by the patient and community" (Wilson-Pepper 1982). Decentralisation of decisions about the provision of health care may result in approaches more consistent with local preferences, and in turn higher levels of utilisation and lower unit costs.

Box 1: Education and Preferences It has been predicted that greater efficiency will flow from a better match between individual preferences and educational services in a decentralised system (Winkler 1989). This, in turn, requires that people are able to exercise a degree of choice in the type of local school that their children attend by, for example, selecting local school administrators, or by relocating to districts which offer the educational services they prefer. Although there are no studies which determine the extent to which decentralisation alters the nature of educational services offered and the degree to which that meets local preferences (Winkler 1989), the following examples highlight several possibilities. (i) Schools may be organised in a way which is most appropriate for local needs - more intensive in the non-harvest, and breaks over the busy harvest season in rural areas. In the BRAC system, decisions about school hours and breaks are made in village meetings, which is said to contribute to the unusually high attendance rates in BRAC schools. (ii) Given regional variations in labour demand, it may be more efficient to provide different curricula, to better prepare people for the skills needed in the locality. Training in seeds improvement will obviously be more relevant to those in rural areas, than those in the capital city. (iii) The debate on the language of instruction is also relevant here. Decentralisation may lead to teaching in the first local language, to meet local demand. This is controversial however, because it may put those children at a disadvantage for future education and employment opportunities (Lockheed 1990). (So could curricula which focus solely on agricultural skills - this raises the same kind of issues.)

Information "...one of the critical properties of all large-scale systems (is that) the authorities must act in an atmosphere of partial ignorance, considerable misinformation, and consequently of great uncertainty." (Easton 1979) Information is an important factor bearing upon the question of efficiency. When there is insufficient or asymmetrical information, it is difficult for government decision makers to predict the consequences of their decisions, and serious efficiency losses can follow. The probability of disparities between a decision maker's ideas about the effects of some policy and the actual local impact of the decision may be much greater in a centralised context. For example, political scientists have found that when information is transmitted through a hierachially-oriented structure, a selective distortion of information is likely given the desire of subordinates to advance their careers through forwarding favourable, and repressing adverse, reports. Such information constraints may be alleviated to some extent by virtue of having decision makers closer to "where the action is", through the creation of (at least partially) autonomous centres of decision making which function independently of the central authority. Schaffer (1982) saw local participation in planning and decision making in terms of the information needed for development programs, posing the question: " how

can decentralisation be organised so that information, local women and men, and program planners can be effectively brought together?". Principals and Agents More generally, it is noted that decentralisation has been interpreted by some economists as an example of the general principal-agent problem: how to provide necessarily decentralised (to maximise information) agents (local government) with incentives to pursue the principal's (central government's) objectives. Substantive principal agent problems arise when there exist, as between different levels of government, conflicting goals and asymmetric information about local costs and needs. Whilst this perspective provides some useful insights into the economic rationale, advantages and disadvantages of decentralisation, it is not a complete basis for the present analysis. Principal-agent theory addresses the question: Given a hierarchial power structure (i.e. a relationship in which one party has the right to write contracts - which the other party can accept or not), what is the best contract that can be written, in terms of the principal's objectives? In the present case, it is recognised that the agents (local governments) have legitimate goals and objectives which may well diverge from those of the principal (centre). Decentralization is addressed here in the light of the task of this survey: what is the appropriate power structure - 'appropriateness' being defined in terms of the objectives of human development. There are two extreme alternatives - on the one hand the complete centralisation of all decision making authority, and on the other the general devolution of such powers - plus all the possibilities in between. It is difficult to evaluate each of these situations in standard principal-agent terms. It might be useful to extend the theoretical analysis to a more complex multi-principal model - i.e. to account for the principal-type roles of both local constituencies and higher authorities. Alternatively, the complexity of the issue of decentralising tasks and policy decisions suggests that a model of intergovernmental bargaining might be more useful. Cost Per Unit of Service "...government costs increase in direct and geometric connection with centralisation." (Proudhon 1959) Whilst the foregoing section outlined the issues raised by economists who explore issues of efficiency and decentralisation in neo-classical theoretical terms, a survey of the literature and country experience reveals that such concerns are not always regarded as relevant. In practice attention tends to be focused upon the more tangible cost dimension of decentralisation - i.e. does decentralisation reduce the unit cost of providing public goods and services? This often involves countervailing factors, specifically the lower costs of local resources versus economies of scale and the risk of wasteful duplication and overlap.

Low Costs of Local Resources Decentralization may lead to lower unit costs, through simpler procedures and building upon existing local resources, knowledge, technology and institutional capacities. Time could be saved, delays due to conflicts between project staff and beneficiaries reduced or avoided, and local responsibility and oversight exercised, so that costs could be lower relative to those when the same services are provided by the central government. Statements which support this proposition can be found throughout the literature (see, for example, Lisk (1985) p.20; Allen (1987); Uphoff (1988) p. 50). Cost reductions are an important motivation for World Bank projects which seek to enhance community participation (Paul 1988). The processing of information involves costs which depend to some extent upon the distance between the origin of the data and their final use. In the process of transmission, each step entails costs, both the direct costs of the intermediary agencies (salaries, administrative infrastructure etc.), as well as those arising from the distortion of information which may lead to expensive misjudgment (Frenkel 1986). In rural localities which are characterised by low population density and relative isolation, it has been observed that public service delivery is "generally more difficult and expensive" (OECD 1986). However local governments may still be relatively more effective in these circumstances than their central counterparts, depending on the significance of such factors as economies of scale (discussed below).

Box 2: Cheaper Local Resources in Water and Education Water programs have traditionally employed the relatively expensive services of professional engineers and other specialised technicians. A greater degree of reliance upon local personnel and resources under a decentralised system may be more cost effective. In Agua del Pueblo, multi-skilled field personnel, similar in concept to primary health workers, have been employed. High school graduates from rural backgrounds receive six months instruction in surveying, basic hydraulic engineering, system design and construction, health education, community organisation and project administration. Working under the minimal supervision of a civil engineer, they provide technical advice and support as the communities work through the steps of installing their water systems themselves (Cox and Annis 1982). Many national governments provide centralised "hugely expensive in-service training" for teachers (Garret 1989 p.309). Yet there is little, if any, evidence available that such formal training is an effective way of producing better teachers (id.). The BRAC teachers, for example, are not fully trained professionals but better educated, younger villagers who take part in an intensive 12-day training course and receive regular guidance and supervision (UNICEF 1990).

There are some considerations which point the other way however - and indicate that decentralisation, through increased duplication and overlap, can increase costs. In Papua New Guinea, the provincial governments "created new public services that have very high per capita costs" (Rondinelli 1983 p.46). In Indonesia, overlap and duplication of

responsibilities between the agencies of the various levels of government has been held to have resulted in considerable inefficiencies. This was also reported to be a concern in Nigeria. Costs of Scale Economies of scale point in favour of centralisation. A certain minimal area and, therefore, a certain degree of centralisation are necessary prerequisites for the provision of some public services, or at least for their cost-efficient provision. This tends to flow from the better utilisation of fixed facilities. It should also be noted that local decisions which rely solely upon local information may ignore the potential for externalities and economies of scale, which in turn creates wider costs for the society. Yet dis-economies of scale may also exist, where costs rise disproportionately with size. This is attributed to the increasing costs of information processing, professionalisation of the public service, and raised expectations in a centralised system. (Diseconomies have been found to be empirically significant in industrial countries: Hawkins 1976, cited in Frenkel 1986). Studies have found that economies of scale exist primarily in connection with those goods and services that are capital, rather than labour, intensive (Frenkel 1986). However it is not possible to determine whether scale produces economies or diseconomies in an abstract way and without relation to the actual public services provided in a given polity.

Box 3: Economies of Scale in Education The proliferation of education agencies under decentralisation can increase overhead and administrative costs. In Latin American countries where the exercise of educational authority is said to bear a "marked political dimension", the pressure to reward supporters has resulted in the creation of multiple structures. The Ministry of Education in Venezuela had 34 central directorates, and the Mexican SEP 7 under-secretariats, 44 general directorates, 304 directorates and 6 councils (Cassusus, 1990, p.12). Decentralised administration may inhibit the realisation of economies of scale in such functions as curriculum development, examinations, textbook production, staff monitoring and teacher training (Lockheed 1990). On the other hand, school-based curriculum development could have a number of advantages. Teachers' skills and professionalism may be enhanced. Changes can be tailored to fit available resources. And teachers have a particular ability to judge the classroom value of ideas; teacher produced materials tend to have, for example, less start-up problems (Garret 1990). However research findings that local curriculum development results in greater job satisfaction and higher productivity are largely reported from countries where the teaching profession is comparatively wellpaid and highly motivated (Garret 1990). One cannot assume that such conditions exist in developing countries, nor that the attitudes of teachers toward additional workload would not be adverse.

Externalities

"Decisions made by constituent units are invariably minority decisions that impose high external costs on the national majority" (Riker 1964) An important aspect of the efficiency question lies in the presence of externalities. Basic education and health care, for example, exhibit significant positive externalities in the form of benefits to the society and economy from having a healthy, literate and skilled population. The benefits will spillover beyond the local community, both directly (if people are mobile) and indirectly, through a more productive workforce. Education has been found to have negative effects on fertility and child mortality, and a positive impact on political awareness and participation (Armitage and Sabot 1988). In a decentralised system, there will be an undersupply of public services when localities only produce a public good until its marginal costs coincide with the marginal benefits for the locality's own inhabitants. The same issue arises in labour economics, from the reluctance of firms to invest in training of employees who may subsequently leave the firm. The possibility that such spillovers might tend to reduce the amount of services provided by the local community (firm) is a concern in the context of many basic social services - although it appears to be less relevant to water supplies. Externalities may be important for economic infrastructure - spending on road construction and maintenance, for example, will tend to benefit road users living outside the local jurisdiction as well as local residents. In order to counteract the tendency to underinvest in a decentralised system it might therefore be necessary to introduce measures to increase local spending on basic social services and economic infrastructure. These might include fiscal equalisation grants, incentives such as matching grants for revenues collected, and regulations which specify minimum expenditure levels, user fees or tax rates. The purpose of central grants which reflect externalities is to ensure economic efficiency in local spending decisions (Helm and Smith 1987 p.xii). National "product" standards, defining minimum requirements in such areas as education and health, also have an important role to play here. This represents a measure of centralisation which undermines local autonomy in decision making about the quality and quantity of provision. It is nonetheless necessary in order to promote levels of investment in services where excludability is not possible, but the positive externalities are significant. This is also related to the discussion of quality in the following section. Quality of Provision One of the many difficulties faced in investigating the comparative efficiency of centralised versus decentralised provision of public goods and services lies in the possibility that quality may vary under the different alternatives. This has direct repercussions for efficiency. If in education, for example, local provision leads to rising rates of failure, repetition or drop-out, any apparent cost savings might be illusory. Centralised control will at least guarantee certain national minimum standards in such areas as literacy and immunisation. This appears to be especially important where decentralisation exacerbates disparities in resource availability between different

localities, so that the poorer areas are forced to resort to services which are lower in quality, whilst the better-off enjoy improved provision. Thus the issue of quality may have significant implications for inter-district equity (discussed below). Under a decentralised system, professional staff may be relatively more isolated and lack the frequency of contact with colleagues, than would have occurred under centralisation. This has been said to lead to a lower quality of service and provision than would otherwise have been the case (OECD 1986).

Box 4: Quality and Decentralised Education It has been said that decentralisation is "the key that unlocks the potential of schools to improve the quality of education" (Lockheed et al. 1989, p.81). This claim is based on the improved performance of teachers, greater relevance of curricula, and so on. Yet there is evidence that decentralised education can be of lower quality, perhaps because local efforts are more likely to be necessary in poor, rural areas where the central government has failed to provide adequate educational opportunities (Winkler 1989). In a locality in which many parents of school children are themselves uneducated and thus unable to reach informed judgements about the relative costs and benefits of relatively high quality schooling, there is a concern that the quality will be less than socially optimal (Armitage and Sabot 1988). Overall, the evidence about the impact of locally provided schooling on the quality of education is mixed. The Harambee schools of Kenya are of low quality, in terms of inputs (teachers) and outputs; very few (13%) of harambee schools exceed minimum standards (grade 4 attainment levels) compared to almost 80% of central government schools. Mwira (1990) argues that a major effect of the harambee movement in education has been the "polarisation of the secondary school system, differentiating school type by quality"(p.360) (See table 2). But this is not the case in Bangladesh, where 95% of pupils from BRAC primary schools have passed examinations for entry into the fourth class of the official school system, and most of these children make the transition to government schools.

Implementation and Maintenance An important advantage of decentralisation, relative to centralised provision, may lie in the better ability to facilitate the implementation of projects and programs in a wide variety of local circumstances. And in the long term, such investments may be more likely to be maintained and effectively utilised where the responsibility lies with local authorities. This is especially held to be the case where decentralisation enhances local community participation. This stands in contrast (and may be partly a reaction) to the evidence that institutional centralisation creates serious inefficiencies in the maintenance of development projects and programs (Wunsch 1991). This appears to be particularly relevant for rural infrastructure projects. A centrally directed and hierarchial bureaucracy has been found ineffective in undertaking various tasks of rural development, particularly under

conditions of uncertainty and severely limited resources. It has failed to secure beneficiary participation, responsiveness to local needs, adequate maintenance and utilisation of infrastructure, and administrative flexibility. It is argued that this leads to serious inefficiencies in public expenditure. Thus decentralisation may be advantageous in its better and more timely decision making by those who have a more direct interest in sensible decisions, given that they are both their authors and the bearers of the consequences (see section 1.5.1). Information problem associated with multi-level hierarchies was noted above. There may also be a loss of selectivity through increased distance. The farther away a centre of decision and the more encompassing its responsibilities, the more difficult it may be for the decision maker to deal with small problems or issues of a predominantly local character which arise in the course of implementation and maintenance. Similar conclusions emerged from donor project evaluation studies at the beginning of the 1980s (USAID 1982; 1983). Maintenance problems were attributed to the competition for funds at the centre - the relatively less glamorous task of maintenance together with the tendency of officials to be distracted by more nationally visible concerns. Further, it was found that highly centralised project selection, design and management led to the construction of economic infrastructure that did not fit local conditions - use patterns, social realities and economic needs. This in turn resulted in underutilisation and local unwillingness to finance maintenance. And where maintenance was the responsibility of centrally based teams which serve many dispersed communities, there were often be long delays in responding to system failures. It is therefore, at least partly, a technical problem - as to how technical information is obtained, and how technical services are provided in an efficient manner.

In Ghana, "local powerlessness has meant that roads, bridges and buildings await repair, and local improvement projects stand idle for months, awaiting national authorisations and funds which come too late or not at all" (Wunsch 1990 p.64).

To summarise, the main issues which recur in the numerous writings which emphasise the inefficiencies associated with centralisation include: weakness in attaining and maintaining beneficiary participation (Uphoff 1986; Korten 1986); underutilisation and inadequate maintenance (Wunsch 1986); irrelevance to local needs and conditions; and lack of flexibility, adaptability and speed (Rondinelli 1983). These observations all impact upon the relative efficiency of the decentralisation/centralisation alternatives, and suggest that thoroughgoing devolution of decision making and authority will aid the effective implementation of programs and projects for human development. Local Participation in Decision Making "The colonists...at least have a greater interest in coming to a right judgement and will take greater pains to do so, than those whose welfare is very remotely and slightly affected ... If the colonists make bad laws, and select improper persons to conduct their affairs they will generally be the only, always the greatest sufferers....and must bear the

ills which they bring on themselves, until they choose to apply the remedy". (Earl of Durham in his 1839 report on Canada, quoted in Frenkel 1986). Decentralization may alleviate technical problems in design and implementation given the greater relevance to local needs, conditions and available resources. If local authorities are willing and able to engage community support and participation their decisions and activities are likely to be based, to some extent, on the knowledge and desires of the local population. This, in turn, is likely to enhance the appropriateness and continued maintenance of local services and therefore the cost-effectiveness of the initial investment. If decentralisation promotes community participation, the management of operations can improve. Irrigation, for example, requires decisions about on-farm use of water and allocations from the distribution point to the farms. Interaction and consensus among water users at the distribution level is crucial for efficient day-to-day operations (Paul 1989). Thus the degree to which decentralisation alleviates problems in the design and maintenance of development projects appears to be largely linked to the issue of community participation. Whilst there exists an extensive literature on the latter, its relationship with the practical and technical aspects of the decentralisation of government does not tend to be an important express concern. In order to bring the two together, it is necessary to bear in mind the discussion of participation in decision-making in Section 3 below. Where decentralisation does facilitate and encourage local participation, then the benefits of community involvement - improved project design and implementation due to better match with beneficiary needs and better appreciation of local constraints etc. - can follow, and result in greater efficiency in government activities, especially in the long term. For example, the use of participatory methods to build reliable water systems and to motivate the complex behavioural changes needed for health improvements requires effective communication with the village clientele. This involves both technical decisions (what size of pipe) and also the resolution of social conflict (where to put the pipe) (Cox and Annis 1982).

The Malawi self-help Water Project recorded impressive results in terms of numbers of village pipes constructed and local institution building. It reflected community consultation and involvement from the outset, and co-operation between the Department of Community Development and local leaders, the government providing materials and training, and the latter mobilising local contributions of labour. District Development Committees - comprised of local parliamentarians, Party members, and the district medical and technical personnel - are the channel of requests for piped water, and are involved in the design process. A villager is locally selected to become the para-technical trainee, to supervise the construction and maintain the pipes afterwards - s/he remains accountable to the community. Although final responsibility remains vested in the

central department, the significant degree of decentralisation has resulted in a scheme which is effective in meeting local needs and resource availabilities. A 1979 study of 34 Guatemalan village water projects built over the previous 15 years found that village involvement in operating and maintaining systems was essential to keep them working. Where villagers received practical training in operation and maintenance, there were fewer system failures and fewer days without clean water due to breakdowns (Cox and Annis 1982). This is an example where transferring the skills and responsibility for repair to the community proved to be an effective alternative to centrally based teams. A World Bank study found that of 25 completed agricultural projects it had financed, only 12 appeared to be achieving long-term benefits (cited in Uphoff 1988). The projects most likely to demonstrate sustained improvements in productivity were those that strengthened institutional capabilities for beneficiary participation in management.

Nonetheless, here, as elsewhere there is an array of country experience. There are a number of factors, ranging from the purely technical to the highly political, which impact upon the long term issue of maintenance. (See also Obstacles below). According to Paul (1989), the impact of community participation on the maintenance of facilities is mixed successful in some functions of operations management but not in others. He concludes that the main reason for failure was the absence of complementary inputs such as effective training and extension services. Wunsch (1990) argued that centralisation and decentralisation (as conventionally practised) are both flawed, and that the vast majority of decentralisation efforts have failed to resolve the inefficiencies which have hitherto plagued development projects. He noted the following deficiencies: (i) The agendas and priorities of central ministries, which continue to control most program and project resources, have seldom been seriously affected (see Obstacles below); (ii) There has rarely been a notable improvement in local broad-based participation local representatives have been field personnel of central ministries, local members of the dominant party, or from the most powerful caste or class factions (see Participation below); and (iii) Local plans are largely illusory, since localities were never allowed to raise and spend significant amounts of their own money (see Financial Autonomy below). The main point which emerges from this critique is that if "decentralisation" follows the mode of deconcentration, improvements in efficiency through better implementation and maintenance are unlikely to follow. It reiterates the point made elsewhere that only where there is a thoroughgoing devolution of decision making power are the presumed efficiency gains likely to be realised.

Box 5: Water and Sanitation The issues of maintenance and community participation are particularly relevant to water and sanitation infrastructure projects. Recent writings from the World Bank (1987; 1988; 1990) have emphasised the finding that centrally-provided improved water supply systems often do not function. It has been estimated that "one in four water systems is not working at any one time, and that the number of systems being abandoned is approximately equal to the number being commissioned. Where breakdowns force people to use contaminated water for even 2% of the time, this risks undoing the health benefits of drinking clean water through the rest of the year (Cox and Annis 1982). Even if they do function they are apparently often not being used. In Cote d'Ivoire and Kenya, for instance, surveys have shown that only one third of the population reported to have access to improved facilities actually used them" (Briscoe et al 1990 p.116). In North-eastern Thailand, evaluations of a USAID project to provide handpumps found that, after five years, most of the handpumps were not working. The problem was that the villagers "did not want handpumps, which were not considered to be any significant improvement over the commonly used rope and bucket. Standpipes were no closer than their traditional sources and so offered no obvious benefits" (Briscoe 1988 p.10). (The IBRD studies tend to emphasise that sustaining and extending services depends upon mobilising "willingness to pay" - which has been the subject of contingent valuation studies. The main recommendation is for increased user charges).

Local Institutional Capacity A growing body of literature underlines the vital importance of local institutional capacity in the operation and maintenance of development projects and programs. Fifty impact studies commissioned by USAID since 1979 have testified, almost invariably, to the proposition that local organisations must be able to shape development projects to their specific circumstances if such endeavours are to be both successful and sustainable (Harbeson 1990). The institutional capacity of local government depends largely upon such factors as the quality and quantity of staff engaged by local government (this is discussed further in Section 6.2.1 below). There must be a sufficient degree of skills, training and management at the local level to carry out the relevant tasks, and enough funds to finance local priorities. There must be a significant degree of autonomy in decision making processes. And so on. The Comparative Extent of Malfeasance If, under decentralisation and the closer scrutiny of the people local government does become more accountable for its actions, the relative incidence of malfeasance may be reduced. The justification for favouring local government, rather than either deconcentrated or centralised administration, lies partly in "the greater capacity of local voters, compared to central decision makers, to evaluate performance" (Helm and Smith

1987). "Goldfish bowl" effects can set limits on nonfeasance and malfeasance by local authorities, as observed in the education context, regarding teachers (see below). Malfeasance would be less likely under a decentralised system where at least some of the following conditions are met: (i) active participation in, and discussion of, local government decision making by the local population; (ii) open, clear and simple routine accounting procedures and reports; (iii) at least a basic degree of numeracy and literacy in the community; and (iv) some means by which malfeasance can be sanctioned - through, for example, the electoral process and the access of local people to cheap and effective adjudicatory systems to ensure due process and public accountability. Yet problems of accountability do arise at the local level - corruption has been described as "rampant" (Manglesdorf, 1988). This ranges from doctors pilfering medical supplies at the local level and selling them in their private practices (as reported in Ecquador) to reports that local councillors and officials have behaved "irresponsibly" with public resources, being concerned primarily with personal emoluments, jobs and status (Stren 1989). In Africa, local government institutions had a "poor reputation...as the hot-bed of corruption, ineffectiveness and inefficiencies" (Olowu 1990). Hughes (1985) states that "all too frequently abysmal standards of morality and responsibility prevail, and local authorities have at various times been suspended on the ground of maladministration (including Banjul, Freetown, Calcutta and Delhi)". Malfeasance (at any level of government) is a waste of resources. The question to be addressed here is its relative significance at local, as opposed to central, levels of government. Corruption has been found to be a serious problem pervading national administration in a number of developing countries (see HDR 1991). In principle, decentralisation which brings decision making closer to the people, accompanied by established mechanisms of local accountability, can improve the operations of government in this sense. Unfortunately there is little direct evidence on this point, nor any reliable estimates of the comparative quantitative significance of malfeasance at different levels of government. Box 6 presents the only available evidence, which relates to the education sector. Overall it tends to support the proposition that accountability is improved, and malfeasance correspondingly less likely, when decisions are made at the local level.

Box 6: The Comparative Extent of Malfeasance in Education The limited evidence available suggests that accountability in education can be enhanced under a decentralised system. In developing countries where inspection and supervision are the responsibility

of higher authorities, these functions are often poorly executed due to inadequate resources (Lockheed 1990). In Senegal, school visits were limited by lack of transportation: only 28 vehicles to be used by over 600 staff in 41 regional directorates. The scope for malfeasance appears to be higher in such situations, than where the local community (particularly parents and principals) are able to supervise, and have the authority necessary to manage and improve instruction. In the Indian state of Karnataka, the Panchayata Act (1985) has involved people in planning and implementing local activities, through the gram sabha (meeting of all voters in the village). There has been a notable improvement in the attendance of teachers who are now accountable to the local community (Sanwal 1987). In Bangladesh, by way of contrast, where government teachers are centrally recruited, supervised and promoted, there is reported to be a high degree of absenteeism and malfeasance (by teachers). Their union strongly opposed attempts to delegate responsibility for teaching to the local level, allegedly because it feared the greater accountability that would result from such a shift (personal communication from a former secretary of the Ministry of Education, 1991). The evidence on this point is not all one way, however. It is reported from Kenya that the Harambee schools, which are not subject to regular auditing, suffer from continued misappropriations of school funds to corrupt head teachers and their school committees (Mwira 1990 p.354). It is also noted that teacher opposition to local autonomy may be justified. One comparative study of the professional freedoms of teachers in 12 countries concluded that "the fact that the school cannot decide on promotions, appointments, dismissal and the recruitment of staff is a guarantee of professional freedoms" (cited in Davies 1988).

Efficiency - Sectoral Considerations Ultimately the question of relative efficiency is largely empirical - unfortunately there exists little quantitative evidence which directly supports, or refutes, the various competing claims. An extensive search of the literature found only a handful of sectoral studies which rigorously evaluated the comparative cost of centralised versus decentralised provision. The distinguishing characteristics and concerns of each of the relevant sectors are summarised in the following subsections. Education Reduced costs are often put forward as an important rationale for decentralisation of education (Winkler 1989). Highly centralised systems of purchase and distribution of school inputs may be manifestly inefficient (as in Brazil, where pilfering and spoilage of central stores is reportedly rife). The structure of centralised bureaucracies may also be relatively costly. Yet the support for greater decentralisation on the ground of cost savings appears to be more theoretical than empirical. McGinn and Street (1986, cited in Cassusus, 1990) concluded that in the field of education there is only slight evidence that these policies work, and on the contrary, that there are many experiences which demonstrate that decentralising policies do not increase administrative efficiency. Case Studies

The construction costs of locally-built schools tend to be lower, due to cheaper materials and lower costs associated with contract administration (Winkler 1989, p19). BRAC schools hold classes in buildings made of bamboo or mud-brick, and children sit on mats on an dirt floor. However, there is evidence that children in schools with electricity and water learn more than pupils in schools without these facilities, although other improvements in construction have shown little impact on educational achievement (Lockheed 1990). There is only scattered direct empirical evidence on the issues of local financing and cost effectiveness. Jimenez et al (1988) claimed to be the first empirical work exploring the impact of local financing. The study, conducted in the Philippines, found that for given levels of enrolment and quality, schools which rely more heavily on local funding are more efficient (i.e lower cost). Regarding quality, it was found that students at schools which relied more heavily upon local funding attained better 'achievement scores' (see table 3) - although it should be noted that these schools also tended to be located in more affluent and urbanised communities. Local funding, collected via a surcharge on taxes on real property and certain cigarettes, is administered by the local school board. The school administration, local government and parents are represented on these boards. Locally generated funds, while small compared to total expenditures, made an important contribution to non-salary recurrent costs. It was suggested that increased efficiency flowed through the following channels: (i) A more appropriate input mix - Whilst teachers were assigned and paid centrally, local funding allowed schools to achieve a more appropriate balance between personnel and non-personnel costs (when central government spending on the latter had been falling). (ii) Lower personnel costs - through lower salaries and/or fewer non-teaching personnel. (iii) An improved incentive structure for local school administrators, accountable to local parents who are better able to monitor performance than central authorities. On the question of cost reductions in education it is noted that teacher salaries and benefits account for as much as 95% of recurrent government expenditures on primary schools in developing countries (Lockheed 1990). Given that teachers' salaries have been eroded significantly in many of these countries over the past two decades, it is doubtful whether further cost-cutting is appropriate. It is important to note that where cost reductions are achieved locally, this may well be due to a shift in the burden of resource provision - from governments to households through extensive in-kind contributions of the local community. Whilst cutting the overt costs of social services is an attractive proposition for central governments facing serious fiscal constraints, this may also impose severe additional burdens on poor families in times of economic distress.

This can be seen in the Zambian context, where the shares of public expenditure allocated to education, particularly to the primary level, have fallen markedly over

the past decade. Over the same period, local contributions to the direct costs of equipment, materials and maintenance have risen - so that it now exceeds the governmental outlay. It is reported that, in vital areas (such as teacher training and basic instruction materials), parental and teacher self help have virtually replaced the state's contribution (Hopper, 1989). Table 4 illustrates the various kinds of contributions that local communities in Zambia make to primary schools. Thus cases which represent "cost-cutting" through decentralisation may also raise important issues about the ultimate division of responsibilities for the provision of basic social services between the state and the private sector, and the redistributitive role of the central government. It has been suggested that decentralisation of education in debtor countries is directly linked to debt renegotiations with the IMF who recommend, inter alia, cuts in public (and ultimately education) spending (Hevia and Nuez, 1989). Health It is generally held that locally-provided primary health care (PHC) is a more cost effective approach than the provision of large central hospitals (see e.g. HDR 1990; 1991). This follows from cost savings on personnel, more appropriate technology, cheaper and effective treatments, and lower overheads. On the other hand, decentralisation may enlarge the scope for delays, supply problems and malfeasance. The literature reflects that there are factors working both ways - enhancing and diminishing efficiency. On balance, the available evidence appears to favour decentralisation - which is largely due to the association between decentralisation and the pursuit of PHC alternatives. Case Studies

In the slums of Bombay, local community involvement in door-to-door immunisation substantially improved the follow-up of triple and polio vaccination (Kowli et al 1990) (table 6). Costs were reduced when the help of schoolchildren and voluntary agencies was obtained. 200 children were invited to take responsibility for 1200 families (6000 individuals), and proved to be effective agents of change in the community. This is cited as 'a prime example of using local resources effectively without incurring extra cost' (p.171). Under a decentralised system, it may be possible to rely upon less qualified staff and achieve good results. It was found in Tamil Nadu, India, that paraprofessional staff (CHWs) paid as little as a quarter of a junior health professional's salary can be effective if tasks are limited and close supervision is available (Heaver 1984). However, since volunteers require close supervision, it may be more costeffective to implement a tightly managed training-and-visit type system. The latter approach has proved successful in producing behavioural change on a mass scale in situations where field staff and clients are poorly educated and geographically scattered. It concentrates on a few key tasks and on selected

clients for maximum impact (e.g. pregnant women and children) and frequent home visits (Heaver 1984). A study of efficiency in primary health care in Indonesia which compared health centres, sub-centres and community health workers, found that community-based health care was the most cost effective approach (Berman, 1989). Both centres and sub-centres provided a similar type of services, through paramedical staff. Cost curves were constructed for specific health functions (curative care, maternal and child health, and family planning). It was found that community health workers were significantly cheaper than clinic-based care - the average costs of community-based nutrition programs, maternal and child health and family planning services were markedly lower than clinic-based services. This provides empirical support for the argument that there is significant potential for cost savings through the delegation of routine services to community health workers. On the other hand, there was no clear relationship between type of facility and average cost. It is noted that the capital, operating, maintenance and supervision components of costs are quite small - generally less than 10%. The variations in costs were related to both utilisation levels at the different facilities and the use of drugs and supplies. Although there did not appear to be any verifiable difference in efficiency between health centres and sub-centres, the latter were found to be clearly favoured by low-income users. This suggests an equity benefit from the decentralisation of health services to more peripheral units. In comparing MCH/FP in clinics with community based care, community health workers were both more efficient and more equitable (ibid p.321). One quantitative study of a public health program in Ecquador sought to test whether decentralised management was more efficient (Manglesdorf 1988). The rural health program recruited indigenous health workers from isolated rural villages into a two month training course, followed by placement back in their villages. The workers were supposed to be provided with the supervision and medical supplies needed to perform their duties. The study measured the productivity of health workers, in terms of numbers of home and pre-natal visits, village meetings and patients seen. Using multiple regression analysis it was found that decentralisation was not a significant determinant of the number of community meetings or patients seen. There was no effective supervision and logistical support. Supply shortages and delays were more severe under decentralisation. Yet overall productivity and decentralisation appeared to be unrelated. Also, there appeared to be an increase in the amount of maternal-child care under decentralised supervision. Within an eight year period the drop-out rate of the community health workers was 17%. Several factors may have contributed to these mixed results. Interviews identified shortages of supplies as the primary obstacle to performance of duties. This, in turn, could be attributed to the difficulties encountered by the government in financing the project on a large scale. Decentralisation worsened problems in the distribution of medical supplies, in terms of delays and shortages. The CHWs had difficulty in managing the revolving drug funds, partly because the price of pharmaceuticals rose sharply (after the 1982 devaluation). The second most cited problem was the lack of effective supervision. Most of the field supervisors had

no training in the supervision of village health workers, were responsible for a range of duties - from supervision to administration, to direct patient care - and were not themselves accountable for their own performance. The observation that productivity was unrelated to the foregoing problems may reflect the ability of the rural poor to organise successfully to provide essential public goods. Accustomed to delay, most health workers either purchased supplies in the private sector or used medicinal plants. The study concluded that the most efficient degree of decentralisation is not only program specific, but also function specific - and that to the extent that functions are separable, decisions to decentralise should be made on a case-by-case basis. Water and sanitation Decentralization in the water and sanitation sector can lead to cost savings in a variety of ways. It may encourage voluntary contributions of materials and labour services. Community involvement in operation and maintenance can save money by reducing the number of expensive post-construction visits by outside maintenance personnel. Yet extensive local consultations and involvement may be more time-consuming and lead to delays. It is noted that much of the available evidence relates to the transfer of responsibilities to local communities and public utilities, and the use of private contractors, rather than to local governments per se. Case studies

In Karachi, the Orangi Pilot Project had a successful low-cost sanitation program. It involved community participation which was facilitated by the modification of standard engineering technology and implementation procedures. Whereas the prices charged by the Karachi Municipal Corporation for laying sewerage pipes were beyond the means of the residents, if the residents formed a 'lane' (the basic unit of the system, comprising of 30-40 houses), then the Project would assist. The Project's social organisers held meetings to explain the program and its benefits to the lanes. A lane manager (selected by the residents) formally asked for assistance. The Project's technical staff surveyed the lane, and prepared plans and estimates for the lane manager. Initial problems with sub-standard work were overcome through a large extension effort. The project became increasingly involved in technical assistance, rather than motivating and organising the people, since the lanes began to organise themselves. Within five years over 1500 lanes had built sewerage systems. The total cost was considerably less than had the task been left to local authorities. The cost to the community was $1.7 million, and to the project $94 000, inclusive of capital expenditures; it is estimated that the Karachi Municipal Corporation would have spent about $8.5 million on this effort (Hasan 1988). The World Bank regards locally and privately supplied water supply systems as less costly than centrally provided. It compares Kenya, where 70% of the costs of the government-run scheme is spent on salaries, to Cote d'Ivoire where a private company operates most water systems and only 35% of expenditures goes to

salaries. Private sector maintenance - by local mechanics - is recommended (IBRD 1987). On the other hand, there is evidence that decentralised provision of water can be inefficient, in terms of high installation costs, inadequate maintenance and so on. It is also noted that water systems built with extensive community participation can take longer to complete - the organisation and training of village water committees, health education and the instruction of local maintenance personnel all require the time and effort of program personnel.

According to Rondinelli (1990) the inability of the local water authority in Guatemala City to manage the system efficiently increases substantially the costs of water services and reduces its capacity to recover revenue. The authority cannot account for 70% of the water in its distribution system - half of which is said to disappear though leaks, whilst illegal connections siphon off the rest. A primary objective of decentralisation and villagisation in Tanzania was the extension of access of clean water to all (Maro 1990). The government had to ensure that people would not abandon the newly established village communities for lack of water. But although there was a general increase in capital expenditure on rural water supply, it was small compared to the existing needs - by 1987, only 42% of the rural population had access to water services (HDR 1992). Field surveys conducted by Maro (1990) stressed the inadequate maintenance of existing facilities - frequent breakdown of pumps and shortages of spare parts, repair technicians and fuel.

Box 7: The Impact of "Demunicipalisation" Upon Efficiency - A Controversial Example. Colombia provides an interesting example of decentralised responsibility for water and sanitation, through semi-autonomous "decentralised agencies", rather than local government. The underlying rationale was that the municipios were inefficient, providing services on the basis of political/ personal criteria which were too charitable to consumers. The agencies, by way of contrast, were seen as technocratic with the virtues of flexibility and expertise. From the 1960s the agencies expanded in both size and number, and local governments became increasingly marginalised and functionally insignificant. There exist quite different assessments of the Colombian approach. The World Bank regards the program, which attained rural coverage of 80%, as the "best in Latin America", and states that local beneficiaries played a major role in all stages, from initiation to operation and maintenance. An administrative committee includes members of local authorities and village development committees, as well as locally elected leaders. In contrast, Collins (1989) is far more critical. He argues that the "demunicipalisation" of service provision in Colombia was selective. Local government continued to provide services in many of the larger urban centres (see Table 5), so that the pattern which emerged was of "a confusing mass of interwoven institutions and lines of authority with shared finances and responsibilities". In practice, according to Collins (1989) the record of the agencies was disappointing. Decision making was highly centralised, delays frequent, corruption rife, and the agencies remained highly dependent upon central funding. Field research in 1983-4 found that the growing social protest

movements in the country held the decentralised agencies responsible for deficient public service provision - and that some 63% of the movement was directed against the agencies (Collins 1989). Since that time there have been significant changes in the structure of state organisation in Colombia - local government has been entrusted with the task of improving basic services, and the decentralised agencies reformed or abolished.

Housing Decentralization can promote efficiency in meeting priority housing needs. It is widely held that greater efficiency requires a shift in the focus of public policies, away from direct central provision, towards the creation of an 'enabling environment' together with a larger role for local government, self-help and community groups. It has been observed that tapping local entrepreneurship, building materials and other underutilised local resources, will lower costs (UN 1989). Rather than replacing national with local direct provision, decentralisation may facilitate the satisfaction of basic housing needs via the creation of an environment which better facilitates private activity in the housing sector. Zoning and land use regulations, as well as basic infrastructure, may be more appropriately decided upon at the local level. When the public sector does undertake upgrading projects, the role of local authorities can be important in promoting community participation. Unclear expectations of what the authorities would do can cause apprehension among beneficiaries leading to local opposition to housing projects (Paul 1989). Community consultation and participation can help to overcome expensive delays and hurdles to implementation, and result in significant efficiency gains. According to Paul (1989) the potential contribution of community participation, in terms of the savings in time and money arising from the smooth implementation of tasks which are normally prone to delays, is especially relevant to housing projects. On the other hand, the evidence does indicate that local government involvement is not a panacea for the housing sector. It has been argued that even with "genuine innovation in technology, materials, financing and institutional support... self help, mutual help, sites and services, progressive slum upgrading, institution building and wholesaling rather than upgrading ... honesty, bureaucratic efficiency and community participation", government provided housing is still "too expensive" (Annis 1987). Indeed, it appears to be generally accepted that public institutions, at any level, cannot directly meet the need and demand for decent housing in a fiscally sustainable manner. Case Studies

In Lusaka, Zambia, during the four years of a World Bank project "8000 houses were resettled ... yet active collaboration with community leadership kept a potentially volatile situation free from conflict" (Paul 1989 p.23). In Mexico, the urban popular movement has been active and successful in organising the poor. Partly in response to this, FONHAPO, a low-income housing

authority created by the Portillo government in 1981, created a window of housing credit to barrio associations, co-operatives and community groups. This transferred considerable power to local groups by allowing them to solicit their own credit, participate in design, select their own technical assistance and contract directly for construction services, which improved housing quality and reduced malfeasance (Annis 1987). In the Colombian context, it has been argued that inexperienced local builders almost invariably use heavier, more expensive construction materials in greater quantity than required. Most projects are small and cannot take advantage of economies of scale. self-help projects also take twice as long to complete as projects constructed by formal housing institutions - four years instead of two (Solock 1984).

Economic infrastructure Decentralisation can lead to more efficient provision of economic infrastructure in terms of both product design and unit costs, and maintenance. The latter point was explored above (sections 1.3, 1.6). As to construction, the relevant possibilities lie in the use of local construction materials and labour, and the more appropriate selection of the type and location of facilities. Where infrastructure policies are applied generally across-the-board, rather than directed specifically toward particular localities, the resulting provision may be more costly and inefficient than what would have been the case where local conditions were taken directly into account. Infrastructure programs which fail to develop or apply technologies for small scale applications typical of rural areas, and require rural areas to use facility designs that were developed for large-scale urban operations, have imposed higher operating costs onto local residents than they can afford to pay (OECD 1986). Sanwal (1987) stated that "it is only now being recognised that (government) programs for human development require spatial decentralisation and co-ordination... The electricity board can lay transmission lines very efficiently but rural electrification requires the planning of these lines according to village concentrations, and an economic return on the investment is dependent upon power connections taken for village industries." Case studies

One example of cost reductions comes from the Baglung district of Nepal. Local committees working under village council auspices, with little outside help, constructed sixty-two bridges covering the whole district, in five years. Local materials and artisans were utilised, and management and labour were unpaid. It is estimated that each bridge cost only one-quarter of what the government would have spent, and was built three or four times faster (Uphoff in Lewis, 1988). In Kenya, high levels of local participation in the ILO/IBRD Rural Access Roads Program contributed to savings in construction and maintenance costs, as well as

higher implementation rates. Active local involvement encouraged people to donate land for project roads - the amount thereby saved was estimated to equal the cost of constructing about 150 km of additional roads (Lisk 1985). In Liberia in the early 1970s, decentralisation was put forward as the new administrative strategy for promoting rural development. Newly created structures included 'development councils' and the restoration of municipal charters. The local administration was said to have been an effective instrument of development in implementing such development projects as the construction of roads and highways. Between 1967 and 1977, more than 3000 miles of secondary and feeder roads were constructed, more than twice as much as had previously existed. A counter-example, of inefficiencies associated with decentralisation, is found in China's electricity sector (Wirtshafter and Shih, 1990). The post-Mao reform promoting greater decentralisation to local government and enterprises coincided with the drive to increase economic growth. This required corresponding increases in electricity production - but in the mid-1980s there were acute power shortages. When the national government (Economic Planning Commission) froze the funding of centrally-planned power plants at the 1986 level, greater responsibility was placed on individual provinces to fund power facilities. The national EPC and the Ministry of Water Resources and Electric Power have encouraged local officials to build small facilities, by granting greater authority and economic incentives.

Whilst China's hydroelectricity is often cited as an example of appropriate technology, many localities are unable to exploit hydro, given low winter stream flow and rapidly increasing demand for electricity. Local authorities therefore exploited the only remaining option, through the construction of small coal-fired thermal plants. Decentralised development has been successful in providing rural areas with much needed energy. The incentives to expand local capacity via small plants lie in local control, the ability to set own prices (and make and keep profits), and significantly shorter lead times. At the same time however, the above mentioned study found that, in terms of thermal efficiency, capital and operating costs, dispatching capacity and environmental impact, the small plants are inferior to the large plants (see table 6). (In 1989, the State Council appeared to reverse its policy on decentralised power plants, by banning their construction. The directive did not, however, discuss the status of small plants currently in existence or under construction; nor did it offer local governments an alternative source of electricity.)
Occasional Paper 13 - DECENTRALISATION: A SURVEY OF LITERATURE FROM A HUMAN DEVELOPMENT PERSPECTIVE

3. Resource Availability

The Distribution of Revenue Sources and Expenditure Taxation Cost recovery

Financial Autonomy and Central Transfers Rationale for Financial Autonomy Central grants Practice Total Public Resource Mobilisation The Visible Benefits Principle and Voluntary Contributions Inter-governmental Transfers - Impact on Resource Mobilisation Borrowing Empirical evidence on total public resource mobilisation

3. Resource Availability There are basically three, interrelated, themes falling under this heading which are relevant to the present survey. The first concerns the distribution of taxation and expenditure under the centralisation/ decentralisation alternatives. Or, in other words, the assignment problem of 'who collects and gets what'. Another relates to the degree of financial autonomy of local governments versus dependence on central transfers. The final theme raises an efficiency issue, of whether decentralisation increases the total amount of resources mobilised for public expenditures - 'how much is there?'. There exists a substantial body of literature which relates generally to each theme; although the empirical aspects of the third are less often addressed. To summarise the findings of the present section, it is seen that the distribution of revenue sources and expenditure responsibilities between different levels of government is such that local public expenditures tend to exceed own-source revenues. The heavy reliance of local government upon central transfers means that their degree of financial autonomy tends to be quite limited; and hence such authorities may in reality be less independent than first appears. However fiscal imbalances do not necessarily indicate an inappropriate allocation of governmental responsibilities and powers. The goals of fiscal equalisation, principles of 'good' taxation and administrative considerations may all point to a substantial role for central transfers. The task is to structure these transfers such that any adverse incentive and efficiency implications are minimised. Finally, it is found that while the impact of decentralisation on revenue raising is controversial, the most relevant evidence suggests that governmental structure and public resource mobilisation have little to do with each other. The Distribution of Revenue Sources and Expenditure The assignment problem is an important theme in the literature about revenue mobilisation in a decentralised system (see, for example McLure (ed.) 1983). It explores which fiscal instruments are most appropriately allocated to the different levels of government. There are obviously important inter-relationships at work here: the appropriate delegation of revenue authority to local governments will depend, to some extent, on the assignment of expenditure responsibilities.

A highly generalised picture of expenditure responsibilities in developing countries might be painted along the following lines. Local government spending tends to be directed towards the direct provision of certain public goods and services, such as primary education and basic health care. Higher authorities have significant expenditures in such areas as defence and security, transport and communications infrastructure, and 'other', which includes public administration. Overall, developing countries tend to have (relative to industrial countries) small, centralised public sectors (Oates 1985). If it is accepted that the national government has primary responsibility for horizontal fiscal equalisation, economic stabilisation and management, as well as significant expenditure responsibilities in such areas as major economic infrastructure, it follows that the centre should exercise control over the most important (lucrative) and flexible, as well as the main redistributive tax sources. The general principles of 'good' (i.e. efficient) taxation, which are discussed below, also point in this direction. Bahl and Linn (1983) argued that for a given set of expenditures, local governments should, on the grounds of efficiency, distinguish between services having the characteristics of (i) pure public goods - which should be financed by taxes; (ii) excludability - financed by user charges; and (iii) externalities - by intergovernmental transfers. The 1988 World Development Report put forward a similar proposition. However it is very difficult to apply and justify such a simplistic classification in practice (Gandhi 1983); an obvious example being a clean water supply, a public utility which has significant externalities. It should be borne in mind that decisions about the assignment of functions and finances to different levels of government can be appraised from a more politically oriented perspective, and the process regarded as one of allocating authority between different (and possibly conflicting) power centres (Bird 1990). It has been said that state and local governments exist in the form they do and with the functions and finances they have, neither to maximise democratic access to government nor to provide public services as efficiently as possible, but rather as a result of many complex historical and institutional factors. Moreover, it is a "common characteristic of most intergovernmental fiscal systems that there is considerable confusion and obscurity as to who exactly is responsible for what and how precisely various public services are delivered and paid for" (ibid). Below we examine two potential sources of revenue for local government, taxation and user charges, outlining the underlying principles and current practice. Attention is paid to the structure of local systems of taxation and the bases which tend to be relied upon; the use and appropriateness of local cost recovery; and the practical constraints faced in revenue collection. It is noted that there are few effective and broad based tax handles available to governments in developing countries, and that "irrespective of the level (of government) to which these taxes are assigned, often their bases are limited and their collection inefficient". However overall it is found that for various reasons, theoretical and political, local tax bases tend to be more limited and fragmented than those exploited by the centre. Moreover, administrative weaknesses appear to undermine the local tax

effort to an extent greater than that suffered by the centre. The discussion about user charges focuses upon the revenue raising ability and equity implications, and concludes that on efficiency and equity grounds there is a strong case for charging consumers of certain public services at higher levels of provision (e.g. for university education). However it is emphasised that those same considerations mean that the nature of the services provided by local governments, especially in the areas of basic education and primary health care, are inappropriate candidates for cost recovery. It is seen that in practice user charges, levied on a consumption basis, have not been widely utilised at the local level. Taxation Theory The basic underlying theory of taxation is fairly well established, and leads to several general principles of 'good' taxation. The first proposition is that a tax base should be inelastic - in the sense that imposition of the tax on a particular good or activity will not lead to a decrease in consumption of the good or engagement in the activity within the jurisdiction. This leads to the Tiebout-type prediction that mobility places limits on variations between the taxation policies of different local governments - and that tax bases which are highly mobile are more appropriately taxed by a higher level of government, unless localities have uniform rates or if variations reflect local preferences and benefits. There are therefore economic constraints on local taxes, given the reactions or distortions which may result. The local tax jurisdiction can be likened to a very open economy, in which the dis/incentive effects of local taxation together with the assumed mobility of local residents and capital place severe limits upon variations in local rates between localities (Brennan and Buchanan 1980). Another general principle is that tax bases which are unevenly distributed between localities - such as those on natural resources and trade - are inappropriate candidates for decentralisation, because of their repercussions for equity. It has been argued that income redistribution and stabilisation are within the proper province of the national government, and that localities should not seek to further these objectives through such means as corporate and personal income taxes (Musgrave 1983). On the other hand, insofar as the feasible set of tax instruments is crucially determined by the information actually available to tax authorities (Newbery 1988), local governments may enjoy a comparative advantage. Information must be collected, tax liabilities calculated and payment enforced. Decentralisation may permit more detailed local knowledge, aiding more efficient tax administration, especially where accounting systems differ significantly between areas (Helm and Smith 1987 p.xii; refer to the discussion of information in Efficiency above). Moreover the so-called visible benefits principle can be interpreted as a justification for the transfer of revenue powers to those levels of government 'closer' to the people (taxpayers), given the increased willingness of beneficiaries to 'share the burden' where they can actively participate in the design and management of projects, and better hold decision makers to account for their actions.

Thus certain kinds of taxation appear to be more suitable for decentralisation - on such items as real property, vehicles and perhaps retail sales. Taxes on external trade and resource rents, personal and corporate income, all of which are important revenue sources in developing countries, are more appropriately levied by the central government. Practice The following paragraphs describe the nature of local tax systems, in particular their fragmented nature, centrally imposed restrictions and the administrative constraints actually faced. It is found that poor local tax performance can be largely attributed to inadequate effort and capabilities, as well as to the disincentives and restrictions imposed by higher authorities. At the same time, inadequate yield is a marked shortcoming of national taxation systems in most developing countries. Tax evasion and avoidance, as well as political obstacles to increased tax yields, exist at all levels of government. The task here is to assess the comparative shortfall at local, vis-a-vis, national levels. More generally it appears that urban municipalities will often be better able than their rural counterparts to mobilise local resources, given their access to more lucrative taxation bases, the nature of local tax instruments (motor vehicles, sales and business), and greater perceived creditworthiness. Tax bases While local governments exploit a range of revenue sources, the general pattern does tend to conform, to some extent, to the theoretical principles outlined above. The major source, which is almost universally applied, is the property tax. In Asia and Latin America, property taxes on buildings and land tend to be the most frequently used instruments for raising local revenues. It accounts for more than two thirds of local tax revenues in the Philippines, and is important in China, India, Korea and Pakistan. The next most popular taxes are those on motor vehicles and entertainment (Bahl and Linn 1983). The former is said to represent an "excellent (for efficiency and equity reasons) but neglected tax base for urban governments" (Bahl, Holland and Linn 1983). The category of 'all other taxes'- "a motley collection of nuisance taxes which are often costly to collect and comply with" (Bahl and Linn 1983) - contribute a sizeable share to local revenue in some jurisdictions. Yet local governments typically face practical difficulties in revenue raising, particularly in the limitations set nationally and administrative problems. Local governments are often actively discouraged from fully exploiting potential revenue sources by a wide variety of central checks, controls and hindrances (Bird 1990). This may be motivated by economic considerations, as well as by popular sentiments. Generally central governments tend to reserve for themselves the most buoyant and lucrative revenue instruments (although, as noted above, there may be sound efficiency reasons why this should be the case).

In Nigeria, the elasticity of local revenues has been severely limited by the removal of wealthier members of the community from the local tax system; all employed persons, institutions and companies were placed under the tax

jurisdiction of the central government's PAYE or direct assessment, thus leaving local government with the poorest sections of the community as its taxpayers. In 1976-7, the taxpaying population of Oyo State fell by 25%, and most of the remaining taxpayers were poor farmers (Smith 1982). There are many examples of centrally imposed legal restrictions upon local revenue raising. National governments often require explicit prior approval for local tax rates, property revaluations and "almost everything else that affects local budgets" (Bird 1990 p.284). If the power to set the rate is part of the "generally accepted definition of a local tax" (Prud'homme, 1989), then many locally collected taxes would fall outside this definition.

The Brazilian constitution limits the municipalities to two sources of revenue certain urban properties (as defined) and specified services (at certain rates). In Nigeria, central approval is required before any levy can be included in the revenue estimates of a local council (Rondinelli, 1990). In Bangladesh, the list of tax sources used by the zilla parishad (district) is "very short" (Schroeder, 1989). The property transfer tax has its rate set in Dhaka and is administered by the central government. Local autonomy is limited by the requirement that tax rate changes must be approved by the divisional commissioner (a central government official) -and the zilla parishad have no effective autonomy regarding taxation, and no independent ability to increase tax yields.

Administrative Capacity Administrative feasibility places severe limits on the revenue raising capability of all levels of government in developing countries (see Newbery 1988). The practical obstacles which inhibit efficient revenue collection and cause substantial revenue shortfalls are by no means peculiar to local government. The prerequisites of staff competence, well-defined tax legislation, and effective means of enforcement present acute problems at the national level (Goode 1990). The tax system is typically in a state of flux, compliance low and enforcement lax (Newbery 1988). Institutional, political, administrative, structural and cultural constraints characterise the tax systems of developing countries at both national and local levels. The interest here therefore lies in assessing the relative revenue competence of local versus central government. Ideally, we would want some quantitative assessments - of net revenue yields (in terms of potential), staff and skill shortages, and the extent of tax evasion - on a comparative basis for the various levels of government. In practice however, one is forced to rely upon more anecdotal-type evidence. In some respects - due to, for example, the lack of qualified staff outside the national capital, the fragmented nature of local tax bases and limited enforcement jurisdiction - it appears that administrative problems may be relatively worse at the local level, especially in rural and remote localities. In some other respects, local governments may be better vehicles of revenue collection, such as in maintaining current registers of taxpayers and property, and cost recovery.

There is substantial evidence which reflects administrative weakness at the local level. Rondinelli (1990) regards this as the most "immediate and crucial problem". A World Bank (1988) survey of 25 countries concluded that administrative weaknesses meant that only 5-25% of local revenues were raised by property taxes, except in Africa, where reliance on property taxes is generally higher and has grown in real terms. In more than half of the Asian and Latin American countries, the real value of property tax revenues has declined. The greatest erosion was experienced in countries with high rates of inflation (WDR 1988). Even when a simpler head (poll) tax is imposed, the task of maintaining a current census of population may exceed local administrative capacity.

In the Sudan, the local sources of revenue - taxes on 'rain-fed' and irrigated land, animals, date trees, urban property, and 'development' taxes in the South - are not calculated in an informed or graduated manner, and rates are kept very low (Norris, 1983). In Monrovia, Liberia, land adjudication teams did not receive the logistical support necessary for their work, so that the local government (MCC) was never able to collect property taxes (Merlin 1990). In the South Pacific, there is reportedly a reluctance to tax rural, particularly customary land (Larmour, 1983). Where, as in PNG, customary rights are unregistered, the costs of registration may be high relative to tax yield. In Barranquilla, Colombia the municipality collects revenues from less than half of the taxable properties because only three of the ten districts in the city have updated cadastre (Rondinelli, 1986).

Are administrative problems any more severe at the local, as opposed to the central level? In some respects, the practical tasks are more difficult in a decentralised system. The shortage of qualified staff (assessors, collectors and accountants) is often more serious at the local level, which is partly attributed to their comparatively low pay, partly to the relatively unattractive location, and partly to poorer career prospects in local government service. In the Philippines, low property tax yields are due to poor records and lack of qualified local staff (Panganiban, 1982). Alternatively, where local revenue officials are directly tied to higher levels of government, programmes initiated and implemented at the local level may be unsuccessful because it is not in the official's long-term interest to pursue the local reform (Bahl 1984). Tax evasion and avoidance are serious problems in many countries around the world (Mathews 1983). Incentives and opportunities determine the degree to which it occurs. Public attitudes with respect to tax morality and the ability of authorities to take deterrent, and if necessary, punitive action, are also important. Compliance may be enhanced by taxation systems which are disaggregated among a number of localities, each of which is less remote from the individual taxpayer than the centre. In West African countries the names of 'delinquent' village taxpayers often remain on display at the local administrative center - which generally leads shamed relatives to pay the amounts owing in a short time (Local Revenue Administration Project, 1983). It should be noted that the penalties for tax offences which exist at the national level are rarely assessed and enforced,

particularly criminal sanctions (Gordon 1990). The perception that tax avoidance will pass undetected may increase rates of non-compliance. Evasion may be particularly acute at the local level, given the more limited jurisdiction and enforcement powers. Expropriation procedures for failure to pay taxes, for example, are generally governed by higher level legislation, rendering enforcement cumbersome and ultimately less likely. Furthermore, opportunities for non-compliance will be increased if tax units are mobile among local taxation systems. Although there are many examples of cases where ineffective enforcement undermines local revenue raising, no quantitative evidence could be found as to its significance relative to that at the national level.

It is reported that Nigerian local authorities have consistently failed to obtain all their estimated revenues for a given year - some collecting as little as a third of the anticipated amount. In 1973, an estimated 40% of taxpayers in the Mid-West State evaded tax, and 70% in the Western State (Smith 1982). In Kenya, where local authorities tax agricultural produce ('the cess'), there have been serious problems because influential individuals and the parastatal Coffee Board are the prime defaulters in payment. This makes collection and enforcement especially difficult (Wallis 1990). In Zimbabwe, on average, only about 10-15% of the potential yield of the development levy (poll tax) is actually collected by the District Councils. Harriss (1988) concludes from a detailed study of local government revenue in the North Arcot district of South India that the relatively slow growth of real revenue in the context of substantial economic growth reflected substantial evasion of commercial and property taxes. In Colombia only about 30% of Banranquilla's business taxes are actually paid each year, because the municipal government lacks effective enforcement procedures (Rondinelli, 1986). In Senegal only about 20% of taxpayers regularly meet their local tax obligations, and summons generally have to be resorted to (You and Mazurelle, 1987). In some cases the central government collects certain revenues more efficiently, given its larger jurisdiction and enforcement powers. In many West African countries, the patente - a licence fee levied on professions and businesses - is a major source of municipal revenue that is assessed, collected and allocated among municipalities by the central government (Rondinelli 1990).

Tax reform and enforcement at all levels may be susceptible to political obstacles. Influential interest groups may include large landowners, representatives of multinational companies and the military. It is arguable that local administrations are more vulnerable to such pressures. It is reported from Nigeria that local taxation is used as a political weapon - supporters of the governing party are completely exempt, whilst members of opposition parties are over assessed (Smith, 1982). Penalities - e.g. the confiscation and sale of property - may be so drastic that these are opposed (and feared) by local politicians and administrators (Bahl et al 1984).

It is reported from Iloilo City in the Philippines that gross underassessment of property values and numerous tax exemptions have undermined the city's revenue base. This situation is attributed to the strong representation of the propertied classes on the city's decision making bodies. In the collection of business and amusement taxes "the city government is at the mercy of the businessmen's willingness to pay", (who often did not) (Ruland and Sajo 1988). There are some examples of successful efforts in increasing local government revenues. Improving the administration of property taxes is evidently an important means of achieving better local revenue performance. Among other possibilities, taxing the ownership of motor vehicles also appears to be a relatively efficient and equitable local source of revenue.

In Ghana, the revenue yield of the Accra City Council increased substantially over the second half of the 1980s (see table 8). One of the aims of the Accra District Rehabilitation Project (ADRP), undertaken in 1984, was to mobilise municipal revenue. This was done through the revaluation of properties by a restructured Land Valuation Board, and the improvement and computerisation of budgeting and accounting procedures (both with technical assistance from the equivalent U.K. office). There is said to be further potential, for doubling the 1988 income in real terms once the revised property tax system is fully operational and collection performance is further improved (EDI 1989). In Colombia, important changes were made to local revenue sources in the early 1980s. Regarding the property tax, cadastral values were updated by applying an annual 15% increase for each year since the last official revaluation, and then automatically updating each year on the basis of the price index. The latter measure has not been fully implemented however; political factors have kept the rate of indexation below the rate of inflation. The property tax rates were also modified, to provide an element of local discretion in rate-setting. Reforms to the industry and commerce tax established a uniform tax base (based on gross sales) and reduced the variation in rates (WDR 1988). In Delhi, property tax collection was improved through a number of reforms including rebates for timely, and penalties for late, payments, better facilities for making payments, and an improved management information system. In addition, taxpayers had to pay their assessments before being allowed to appeal in court. This resulted in increasing revenue by 16% in 1985-6, and by 96% in 1986-7 (WDR 1988). In Seoul, automobile tax revenue more than doubled between 1982 and 1986, while total local tax revenue increased by 64% (WDR 1988). Traditional structures may facilitate, or limit revenue raising. In practice it is more often observed that the diminution, or abolition of 'native authorities' has caused difficulties in local tax collection. In Zimbabwe, prior to independence, rural local authorities had a reasonably firm local revenue base because of the participation of traditional chiefs and their subordinates in the system. Since traditional leaders have lost their land allocation and judicial powers there has been a sharp decline in the yield of the development

levy (formerly the poll tax), whose collection was transferred to the district councils (Mutizwa-Mangiza, 1990). Some District Councils have reassigned tax collection powers back to the traditional leaders (see case study). Similarly, in the Sudan, collection of rural sources of revenue depended on the knowledge and influence of traditional heads, the odmas and chiefs in the north and chiefs in the south - and both odmas and chiefs were abolished or compromised after 1969 (Norris, 1983). In 1985, a comprehensive strategy of revival of the old system of native leaders was initiated. The intention was to allow local chiefs to participate in and contribute to the administration of the country, for the sake of achieving political stability and efficiency in revenue collection (Elhussein 1989).

Cost recovery Local and central authorities may seek to recover the marginal or average costs of publicly provided services, including depreciation and interest charges. This possibility gained increased attention over the last decade as the international monetary agencies explored means of increasing revenue mobilisation in developing countries (see Jimenez 1988). There are, however, very significant concerns surrounding the appropriateness of such charges, particularly for basic social services. User fees for basic social services raise important efficiency issues, particularly where significant positive externalities exist. A health clinic, for example, provides benefits to those not inoculated against contagious diseases. Efficiency requires charging users less than the marginal costs of the vaccination service, and financing the remainder of the costs from general revenues. If the price is set too high, the service will be underused and the society as a whole will suffer from not having a healthy and (at least potentially) productive population. Similarly, the social rates of return are high on investments in all forms of education, particularly primary education. The benefits are also indirect (higher productivity) and non-monetary (lower fertility rates). Further, distortions in related markets (credit and insurance), which can be viewed as a divergence between private and social benefits, have important implications for efficiency. There are also significant equity concerns. Take the example of clean water, and the facts that the income elasticity of demand is relatively low, and that household size and income may be negatively correlated. Recent evidence drawn from Ghana, Lesotho, and elsewhere has shown that the introduction of user charges led to decreased utilisation of the relevant social services, particularly in rural and remote areas. Although negative repercussions might be avoided through different charges for different income groups, means testing is administratively difficult. On the other hand, it has been argued that blanket policies in favour of 'free' provision across the board have resulted in inefficiency and inequity (Jimenez 1988). It is said that the absence of fees has led to underinvestment in the social sectors and the inability to ration scarce services according to need. Access to free public services is limited in practice, and those who are forced to resort to private provision are often the poorest. The

better off, who have been found to enjoy greater access to the most costly services (e.g. university education and health care in urban hospitals), have been subsidised the most heavily. It follows that cost recovery must be sufficiently high to provide adequate revenues for maintenance of an effective service, deter over-use, and avoid rationing by less efficient and less equitable means. This argument is more convincing with respect to higher level, rather than basic, services. According to its advocates, user charges are particularly relevant to local government - as a means of financing recurrent costs and avoiding reliance upon central transfers (e.g. Bahl and Linn 1983). Indeed, the first of the four criteria for efficiency in raising local revenue listed by the World Development Report 1988 is that "the cost of providing local services should be recovered, to the extent possible, from charges on the beneficiaries" (p.159). This is said to be "especially important at the local level because, being closer to beneficiaries, local public services are more amendable to such charges than services provided by higher levels of government" (p.160). The World Bank (1988) reported that in 25 developing countries surveyed, user charges accounted for nearly one-third of all locally raised revenue. Within its definition, however, the Bank includes both consumption-related and benefit-related charges. The latter are one-off payments for a particular project having the characteristics of both taxation and user charges; although given that one cannot choose whether or not to use the service (pay the levy), it is more akin to a tax. Betterment levies have been used extensively in Colombia, India and Indonesia as an indirect means of recovering costs from landowners pro rata who enjoy the benefit of public improvements nearby or on their property (Bird, 1984). In New Delhi, for example, such charges are used to recover 50% of the cost of public works (Rondinelli, 1990). Alternatively, municipal governments have recovered the cost of services through land readjustment - whereby landowners pool their property for service improvements and contribute sufficient land to compensate the government. This has been done successfully in Taiwan, Thailand, South Korea (Rondinelli, 1990). Clearly, municipal governments must have the requisite technical capability to levy and collect such charges. It has been argued, for example, that metered water is a "pipe dream" in most developing countries (Gandhi 1983). Betterment levies have met difficulties in assessing the surplus value attributable to public improvements, and determining the "zone of influence" in identifying the properties benefited. It may also be hard to estimate total costs prior to project completion. This was a problem in Colombia, so that collections tended to be unstable and erratic (Bird, 1984).

In the early 1980s, it was reported from Nigeria that the significance of local user charges, which were largely confined to urban townships, had tended to decline. In the Northern Region in the mid 1960s they contributed 15% of native authority revenues, but by 1971 local authorities in the North Western State only raised 1.5% of revenues from this source (Smith 1982).

In Jakarta, delayed notification of the Tax Department that an area would be improved and hence the late notification of landowners, together with difficulties in tracing landowners because of unregistered transactions and a lack of data on land values, all created problems for the administration of the city's betterment tax (WDR 1988).

It may be possible to tap non-cash sources of revenue - in the Malaita Province in the Solomon Islands, hospital outpatients are required to bring potatoes for the canteen (Larmour, 1983). Levying charges may be politically difficult. In Pakistan, for example, local authorities were reportedly unable to recover costs from squatter communities, even though (or perhaps because) some of the residents were not poor (Rondinelli, 1990). National governments may limit the utilisation of user charges on the grounds of inflationary pressures, or feared political repercussions. Thus the issue of cost recovery is, as the discussion in HDR 1991 indicates, a complex one. It is submitted that there may be damaging repercussions for the principle of equity and universal access, and the net revenue gained may not, in any case, be significant. Whilst this basic argument applies to a range of public goods and services, it is particularly relevant to priority social expenditures. The Report concludes that primary schooling should be free, and that there is a strong case for secondary education to be free also; that primary health care should be free, although some charges for drugs and hospitals may be justified; and that users might contribute to the recurrent costs of water and sanitation, whilst the government should bear the capital costs (p.67). At the same time user charges could be introduced which discriminate on the basis of type of service, exempting, for example, residential water, basic education and primary health care; and recovering a substantial part of the costs of such higher level services as sophisticated health care in urban hospitals and university education. This implies that central governments are more likely to be able to levy fees with fewer adverse repercussions, given that the level of provision is generally higher; and that local governments, which typically provide basic social services, ought to rely on central grants and local taxation rather than user charges. The review of country experience showed that, apart from the special case of betterment levies, local authorities in developing countries have rarely been willing and able to fully exploit user charges as a revenue instrument. Zimbabwe, Lesotho, Kenya, Indonesia and Botswana have abolished primary-level education fees during the past decade (Jimenez 1988). Financial Autonomy and Central Transfers It follows from the foregoing discussion of local revenue raising possibilities that vertical fiscal gaps are likely to emerge, and that intergovernmental transfers will play a critical role. Here we examine intergovernmental financing arrangements, with particular interest being accorded to the impact of the structure of transfers upon the degree of local autonomy. (The significant redistributive aspects of such arrangements are discussed

below, in the context of Equity, and are not detailed here; the impact of grants on total resource mobilisation are discussed in section 2.3 below.) First this section explores the basis for the widespread belief in the need for financial autonomy at the local level, at both theoretical and practical levels. It goes on to describe the nature of the systems of central grants generally found in developing countries. While cognisant of the difficulties of measurement, it proceeds to present a picture of the limited degree of local government financial autonomy which exists today. Rationale for Financial Autonomy Financial autonomy can be defined as the ability of government units to finance own expenditures from own revenues. Thus in the model presented above, it was presented as a special case of the general allocation of taxes and expenditures. One recurring theme in the literature on decentralisation is the importance of financial autonomy. This is invariably seen as a crucial ingredient in explaining the success (or, more often failure) of programmes for decentralisation. As one well known observer noted, lack of financial autonomy "reduces the burden, but increases the dependence of, local authorities, who generally neither impose taxes nor have to justify to local populations how money is spent" (Rondinelli 1983 p.49). To summarise, the emphasis on local financial autonomy is based on the following propositions: (i) Independence. It is generally believed that genuinely independent decision making will not emerge at the local level unless local governments have a significant degree of financial autonomy. Otherwise local authorities will merely be the 'deconcentrated agents' of the center. (ii) Incentives. Local decision makers lack incentives to behave 'properly', and according to local needs, unless the resources being expended are those of the local constituency, to which the decision makers are accountable. In other words, financial autonomy makes local decision makers more sensitive to both costs and local priorities. On the other hand of course, equity considerations may well point in the opposite direction, in favour of central transfers. Once the national government accepts the role of ensuring at least a minimum degree of parity in the provision of public goods and services among localities, some form of fiscal equalisation grants would tend to follow. Given the substantial regional disparities which characterise many developing countries, this redistributive function is likely to be crucial. Grants are also a primary means for overcoming the problem of spillovers (externalities). Further, as to incentives and accountability, there exist other perhaps more effective means of ensuring that local decision makers are responsive to both costs and local priorities, such as local elections. The nature of the grant system is another crucial variable - strictly tied grants have more far reaching repercussions for financial autonomy than do lump sum transfers. These

points are often forgotten by those commentators who lament the lack of financial autonomy. The available quantitative evidence as to the extent of local financial autonomy shows the following. For the small sample of countries for which data is presented in the IMF GFS, we can observe a range of degrees of local financial autonomy (table 1). Measured in terms of local own, as a percentage of total local revenue, Indonesia is the least (at 22%) and Kenya the most independent (98%). Local own revenue as a proportion of local expenditure shows a similar pattern. The average degree of local autonomy across the sample stands at around 64%. Table 1 Local Financial Autonomy in Selected Developing Countries, 1988 Own revenue as a % total local revenue Chile Colombia Indonesia India Kenya Malawi (1984) Malaysia South Africa Thailand Zimbabwe Brazil 64 87 22 56 98 95 73.6 88.7 68 51 79 Own revenue as a % of local exp. 61 59 22 48.8 98.9 67.4 64.8 90 73.6 58 75.9

Experience has shown that financial dependence can increase local government vulnerability to central domination, in terms of formal controls (e.g. tied grants) and more covert modes of influence, as well as the economic fortunes of the national government.

In Nigeria, the objectives of the local government system introduced in the 1970s - to make local administration more responsive to the local community - were found to be incompatible with its revenue base (Smith, 1982). Local governments are heavily dependent upon central grants, which can be withheld if local expenditure is judged "wasteful, unreasonable or in conflict with central policies" (Smith 1982). The author reports that "these controls have tended to be drawn into a punitive style of central supervision motivated by partisan interest as much as a concern for fiscal irregularity". It was also noted that 99.6% of grants for education and health were in arrears, causing severe cash flow problems (in 196970). The heavy dependence of state and local governments on the monthly federal allocations had further repercussions. The national dependence on oil led to extreme instability, as both prices and production oscillated on a monthly basis,

and made the planning of government expenditures, at all levels, virtually impossible. In Tanzania, the national government provides 100% grants for expenditures on education, water supplies and roads - and correspondingly insists on some significant controls over local policies and expenditure. But the most significant problem facing Tanzanian local authorities since 1982 has been the general decline in central transfers, itself a result of macro-economic constraints (particularly the rising burden imposed by debt servicing). At the same time, local dependence on central transfers has increased significantly, from 60% (1984/5) to 77% in 1987/90. These trends have diminished the ability of local authorities to run key social and economic services, and to maintain local infrastructure (Mutizwa-Mangiza 1990). The author draws the negative lesson that it is "very unlikely that much decentralisation can be achieved in circumstances where central government accounts for much of local expenditure" (ibid p.30). In Bangladesh, actual grant allocations tend to be erratic. Bahl (1989) found that the average shortfall is just over 20% - which is a serious problem for local governments with very little own revenue.

There are a number of instances where the central government has interfered in local government finances in an apparently arbitrary and disruptive manner. In Nigeria, for example, the abolition of the cattle tax in 1974 and the community rate in 1980-2 was done unilaterally, and without compensation. These items had constituted 50-90% of local internal revenue at the time (Orewa, 1987). Central grants In principle, central grants can have several objectives. These include the reduction of vertical and horizontal fiscal imbalances; reduction in differences in fiscal capacity between local governments resulting from different resource endowments and economic structures; ensuring a similar pattern in the provision of services in each locality, at least to a minimum level; and internalising externalities - the use of grants in the latter case derives from Pigou (1947) and is usually referred to as "the compensation principle". According to Bird (1990, p.281), the "essence of the (intergovernmental) finance problem is how to adjust fiscal transfers to achieve tolerable results in the face of clearly nonoptimal assignment of functions and finances". Grants can be differentiated by the degree to which they are need-related (Bennet 1982). Grants can be allocated on the basis of measured geographical differences in needs, or in costs of servicing those needs. These arise mainly from differences in the number of client groups, and in the size and density of local jurisdiction. Usually expenditure need is measured on the basis of a national minimum standard. Alternatively, grants can seek resource equalisation in the revenue base of local governments, given the same tax effort. Revenue-equalising grants seek to overcome unequal tax bases by distributing grants preferentially to those local authorities which for a given tax effort have a small tax base as against those authorities with large tax base. Unitary grants combine features of each

of the two approaches, with the advantage of simultaneously compensating for disparities in both tax bases and local spending needs. More generally, central grants can be either block or categorical in form. General purpose block grants (lump sum transfers) are more efficient in the sense that local governments can decide how best to spend the money. In practice however, spending may still be subject to central guidelines which effectively govern allocations. Categorical or projectspecific grants are intended to be used only for specific categories of expenditure - either broadly or narrowly defined. The allocation of categorical grants can be based on a formula (such as indicators of 'need' - e.g. number of primary school children) and/or a project-by-project review of specific proposals. These may have the advantage that the center could direct funds to areas of priority for human development (assuming, of course, that the center would, and that local government would not, have done so). Funds thereby directed to priority areas may be additional, or alternatively, this may simply substitute for what would have been spent in any case (since central grants, like overseas aid, is fungible). Grants can be adjusted by matching requirements, which aim to induce a degree of local involvement and accountability for the particular expenditure programme. Grants can also be distinguished on the basis of whether they are open or closed-ended, i.e. whether the central government will finance all eligible requirements or whether an upper limit is set.

In Bangladesh, the most important grant revenue source of the zilla parishad (district) is the Upazila Development grant (which replaced the Rural Works Programme in 1984) (Bahl, in Schroeder, 1989). It is a block grant, insofar as there is considerable local discretion in project selection; however there are significant elements of central direction, through the establishment of guidelines and priorities. It does not have any matching provisions, nor any other incentive for local governments to increase their revenue effort. A second type of grant to rural local governments is known as a normal grant - which is in fact a collection of specific purpose grants, earmarked for various purposes.

Where centralised planning is important, transfers are likely to be plan-related and consequently earmarked for development projects approved thereunder. It has been observed in Indonesia, Malaysia and Thailand that the heavy dependence of local authorities on central transfers and the lack of financial discretion in expenditures has weakened local government (Nooi, 1987). In Indonesia, budgets of the regional authorities require central Ministerial approval, while those of the lower levels are subject to action by the executive at the next highest level. In the Sudan after 1979 powers and functions were channelled down to the most basic local government unit, but this was not paralleled by a corresponding decentralisation of personnel or financial powers. In some cases (e.g. Nigeria) the formula for revenue sharing is provided for in the constitution. More often however, it has been observed that transfers tend to be regarded by the center as a residual in their own budgeting process (Bahl and Linn 1983). Obviously the extent to which this occurs depends upon the seriousness with which local

institutions are regarded by higher authorities. It may also reflect, in part, the stabilisation objectives of the national government. In 1979, only about half of the grant entitlements of local governments in the Philippines were actually distributed. Grants should normally, as far as possible, exhibit the features of stability, certainty and objectivity, in order to promote efficiency and allow appraisal against social, economic and political goals (Bennet 1982). However the system of intergovernmental transfers in developing countries is often closed, complex, and subject to uncertainty and political manipulation. In Brazil, for example, 26% of federal transfers to the states and local agencies are governed by "convenios" - specific agreements negotiated on a discretionary basis, in the tradition of "clientismo". Practice In practice, the conditions for complete financial autonomy are rarely met. Virtually everywhere, expenditure exceeds revenue at the local level (see table 9 above), which portrays the practical significance of central transfers. It is interesting to note that vertical fiscal imbalance is not confined to developing countries; indeed a recent comparative study found that central transfers account for a greater share of state and local revenue in industrial countries than in developing countries (Wasylenko 1987). There is a wide range of experience - local own revenue as a proportion of local expenditure ranges from 100% in Mexico City (Bahl and Linn 1983), to about 15% for Zimbabwe's District Councils (Helmsing 1991). The sample of developing countries drawn from the 1990 IMF GFS shows that the degree of local financial autonomy typically fell within the 50-90% range, and averaged around 64% or more. An alternative approach is to observe that the average share of local taxes in total revenue rarely exceeds 20% (and is often much less in many African countries); whilst the share of local in total expenditures is often higher than 30% (Prud'homme 1988) - this implies that about one-third of local expenditure will be financed from central transfers. In a number of cases central transfers constitute a significant source of local revenue.

In Malaysia the national government provides more than 85% of total local revenues, in Pakistan, 89%, and in Indonesia, the figure is about 75% (Rondinelli, 1990). In Nigeria, local sources as a share of total revenue dropped from 75% in 1962, to about 19% in 1983 (Olowu, 1987). In Zimbabwe, the contribution of local revenue to total district revenue increased more than three-fold between 1980 and 1985, yet it still only averages about 15% (Mutizwa-Mangiza, 1990). In Tanzania, the share of local revenue in total local government revenues declined from 40% in 1984/5 to 23% in 1987/8 (Mutizwa-Mangiza, 1990b). In the Sudan, the share of local revenue varied between provinces, from 29% to 10% (Norris, 1983).

The extent to which central transfers constrain local autonomy depends largely upon the terms of the allocation - whether block or conditional, whether there are any accompanying guidelines and so on. This has to be examined on a case-by-case basis. In Indonesia, for example, under the Inpres programme for schools, local discretion has been limited to location, and decisions as to how many, what capacity, design etc. are resolved by the center. Similarly in Zimbabwe, central grants are strictly tied to specific purposes. Total Public Resource Mobilisation In many developing countries, increasing public expenditures on human development requires more than restructuring existing budgets. There may also need to be an increase in the public revenue and expenditure ratios (HDR 1991). Tax revenue as a percentage of GNP stands, on average, at 15.5% for developing countries, compared to 23.2% for industrial countries (HDR 1991 table 44). Similarly developing countries, on average, have smaller public expenditure ratios than industrial countries - 20%, compared to 28% (WDR 1990 Table 11). It is also noted that public sectors in industrial countries tend to be substantially more decentralised than developing countries' (as reflected in the structure of public employment; see table 10). It is therefore of interest to explore the possibility that decentralisation, through the creation of effective lower levels of government, may be a means by which the need for increased total public resource mobilisation can be met. The following paragraphs examine the so-called visible benefits principle, which is generally the basis for optimistic predictions about the revenue raising potential of local government and the significance of voluntary contributions at the local level. (The reader should also bear in mind the discussion of local taxation and cost recovery above.) The impact of central transfers upon local tax effort is also relevant. Further it is possible that local borrowing constitutes a significant source of local resources. The mechanisms by which local resources can be mobilised raise several, often competing, considerations including revenue raising ability, effects on economic efficiency, equity implications and administrative feasibility (Bahl 1984). Finally it is useful to seek some overall empirical indication of the impact of decentralisation upon total resource mobilisation. The Visible Benefits Principle and Voluntary Contributions The visible benefits principle of taxation predicts that decentralisation will elicit additional contributions, even among the poor, and from the poorer regions in developing countries. Decentralisation may lead to greater mobilisation of local resources through the willingness of consumers of public services to "share the burden" when they can actively participate in the design and operation of projects. This would seem to apply only where local residents believe that a significant proportion of taxes will be retained locally, and spent on projects and services which they value.

Evidence to support the visible benefits principle is provided by a detailed survey of development resources and constraints in one of the better administered

upazilas in Bangladesh. It found that among those people who indicated their readiness to contribute to local development activities (about half the respondents), the poorer groups were more willing to do so than the better off, provided that they did not have to carry a disproportionate share of the burden (Thomas 1989). According to a study of decentralisation in Mozambique, the decentralised health service encouraged people to express their preferences and correspondingly adapted to local needs - to a greater extent than other social services. And as a result, health benefited more than other sectors from an increased input of locally generated resources (Mackintosh and Whyt, 1988).

The visible benefits principle can, of course, work both ways. As we have seen, tax evasion is a serious problem at the local level. Non-compliance may be attributed to perceived lack of benefit. According to Rondinelli (1990), the widespread image of inefficient municipal management explains the African city dwellers' reluctance to pay taxes. In Nigeria, the general public is said to be strongly resistant to property rate collection and other charges - a hostility due in part, it is said, to the fact that residents do not feel that they are getting the amount and quality of services for which they are paying (Rondinelli, 1990). An evaluation of the Monrovia City Council in Liberia stressed that people do not pay property taxes, nor fees for services, if services are poor, enforcement is weak and funds are wasted (Werlin 1990). In Zimbabwe the very low yield of the development levy is due, in part, to the perception that the funds will be "wasted" on local council administration (see case study). Indeed it has been observed that (former) Prime Minister Thatcher's approach to local government in the U.K. was based upon a belief in a visible costs principle - that people will actually not demand local public services once it is clear that they must directly contribute to their provision. The visible benefits principle must be viewed in the context of taxation - principles and practice - discussed above. The reader will recall that the revenue raised at the local level is often quite limited. Perhaps the practical significance of the visible benefits principles lies in the local mobilisation of voluntarism, which taps resources for public investment which might otherwise have remained untapped. The revenue raising potential for local development can be considerable, as evidenced by the Harambee movement in Kenya and the Tesito in the Gambia. It has been argued that successful decentralisation is a crucial complement to effective self help (Schaffer, 1982). The latter "requires marketing and credit facilities, water and infrastructure and so on, which are likely to depend on the decentralised institutions of government" (ibid. p.42.)

The Orangi project in a Karachi slum area, which focused on women's health and education, and sanitation, mobilised self-help activities to co-finance and construct water, sanitation and other facilities, and was supported by outside technical assistance (from government, NGOs and volunteer social workers) (Rondinelli, 1990). In the Indian state of Karnataka, the gram sabha (meeting of all voters in the village) assists in the implementation of local development schemes through

volunteer labour and the collection of local monetary contributions (Slater and Watson, 1989). In Kenya, the late President Kenyatta actively encouraged the harambee movement (Swahili, meaning 'let's pull together'). County councils played a vital role in the harambee movement through Community Development Assistants, whose task it was to help local groups organise themselves to run projects in areas such as agriculture, education and health. They were expected to play a role as links between the community and government (both local and central). By way of contrast, self-help is said to have been ineffective elsewhere, for a variety of reasons, due mainly to the political and bureaucratic context where official attitudes and policies discourage such activities.

In Calcutta, because the Calcutta Metropolitan Development Authority "has little communication with the communities in the metropolis", self-help activities have not been effectively mobilised (Banjaree, 1989). Mobilising enthusiasm for local self-help projects has been largely unsuccessful in community government areas in the North Solomons Province of Papua New Guinea. Most local leaders are said to look to provincial grants in order to launch community development projects (Nyamekye, 1983). District Councils in Zimbabwe have reported increasing disillusion with respect to self help projects, largely because the central government promised contribution frequently fails to eventuate - e.g. locally constructed schools stand empty for want of teachers (Annual Plan Reviews, various). Some authoritarian governments effectively prohibited community self help organisations (e.g Argentina, 1976-1988). Factors identified as key to the successful mobilisation of voluntarism (Khan and Miller 1989), include such elements of the local context as culture, religion, the political climate, government policies, and the availability of wealth, labour and management skills, and time.

Voluntarism has been explored in the Bangladeshi context (Khan and Miller, in Schroeder, 1989). It was observed that in Islamic and Hindu doctrine, there are established concepts relating to contributions towards some kinds of collective goods. In Islam, the concept of zakat (one of the five pillars of the religion) is translated as alms giving. There is strong evidence that local communities are able to mobilise funds on a regular basis to support local religious institutions. Chanda (Bengali, meaning dues paid to a specific cause) is frequently collected in cash or kind for religious purposes, and is also a financing source for secular projects such as road repair. It is collected from the direct beneficiaries of a particular project. Fieldwork has found that the significance of such contributions varies between districts. In some countries university students who receive government subsidies are required to "volunteer" to help implement and supervise small-scale rural development projects - Nepal, India and Indonesia provide examples of this.

There are widely used terms for the donation of labour in countries as diverse as Sri Lanka (shramdan) and Peru (minga). The efficiency of "voluntary" contributions of labour and materials depends upon the opportunity costs therein. At certain times of the year (such as the non-harvest season), the opportunity cost of labour may be extremely low. Efficiency requires, however, that the society-wide benefits of the project at least equal the opportunity costs of the individuals who donate labour and materials. Moreover, in-kind services raise important questions of equity. Local leaders and traditions may ensure compliance with such contributions, but the burden may not be equitably distributed. It is also noted that the use of compulsory labour on local projects by colonial regimes left a negative legacy toward such a policy. Still local responsibility for the construction of buildings with local labour, to house services run with central government staff, is characteristic of many parts of francophone Africa (Bahl, et al 1984). Some are more sceptical about the scope for increased resource mobilisation via the voluntarist route. Bahl et al (1984) argued that voluntary contributions, both monetary and in-kind, rarely account for substantial amounts of revenue, and that their growth potential is limited. Moreover, 'self-help' associations may become a drain upon central government resources. They tend to focus more upon the construction of facilities, rather than the maintenance thereof - and as such the national government may be under pressure to meet recurrent costs (Hughes 1985 p.62). In Kenya, community groups depend heavily upon central government financial support for capital and especially recurrent expenditures. Communities may embark on the construction of health centres, for example, on the assumption that the government would subsequently supply drugs and equipment, and employ the required staff. The community development assistants, together with local politicians, have tended to urge communities to embark on more and more projects without consideration of longer-term factors, such as likely recurrent costs. Wallis (1990) reports that central government budgets thereby came under "excessive pressure" (p.440). This resulted in the assertion of greater control by the central government, and the reduction of local authority involvement. Inter-governmental Transfers - Impact on Resource Mobilisation "All assistance should be provided in ways and on terms that are positive-sum - that neither substitute for nor discourage people's contributions." - (Uphoff 1988). As discussed above, transfers from higher to lower levels of government are often a significant source of local finances. Local resource mobilisation may either expand or contract in response to the transfer. In the former case the transfer can be termed "stimulative", whilst in the latter local governments cut back on local revenue raising, and substitute the central transfer (Bahl et al 1984). The objective of central transfers should be to allow recipient local governments to fulfil their expenditure functions, while encouraging their own fiscal efforts as far as possible and ensuring an equitable distribution of public service provision throughout the country.

According to Bird (1990), an increase in total resource mobilisation is unlikely to follow central grants in practice. Similarly, the World Bank (1988) expressed the view that local governments come to view grants as substitutes for local taxes and user charges (pp. 1656).

In Nigeria in the mid-1970s, grants from central and state governments to the local level were substantially increased. From 1976 to 1977?, central grants to local government increased roughly five and a half times. At the same time, the national revenue sharing formula lacked any incentive for raising taxes; as a result local governments were implicitly encouraged to drop such revenue sources as community taxes, property rates and user fees (Oluwu 1982). Both state and local revenue fell, in relative and absolute terms (Olowu 1990). In Ibadan City, for example, the number of property taxpayers dropped from 27 000 persons in 1975 to 8 650 in 1979. Other cities experienced similar declines (WDR 1988). In Bangladesh, some components of the 'normal' (tied) grants programme are supposed to provide an incentive for local resource mobilisation. For example, central grants are used to pay a portion of the salary of local officials. But in practice this does not stimulate local revenue effort - the grants are not conditional, and the local salary share often goes unpaid (Bahl 1984).

Matching requirements can be used to stimulate local resource mobilisation. Under these conditions, it is in the locality's best interest to continue raising revenues which, in addition, ensure that it has a more vested interest in the overall success of the project. The same arguments would apply to the use of distribution formulas for block grants which specifically include local tax effort. There are examples of specific attempts to provide incentives for local revenue raising. The state of West Bengal enacted legislation in 1983 which requires municipal governments to increase their own revenue collections as a condition of capital assistance grants (Banjaree, 1989). (Grant formulas in the UK, Japan and some state grant programmes in the US include a reward for local tax effort: the grant is calculated as the difference between a jurisdiction's current property tax collections, and the amount it would have collected had it applied its existing rate to an 'average' tax base: Dillinger in EDI 1989. Borrowing Borrowing is a means by which additional funds may be mobilised for local governments. National governments in developing countries have increasingly allowed municipalities to borrow to finance infrastructure, especially when fees can be levied to recover costs (Rondinelli 1990). At the outset however, it should be pointed out that this may be a problematic route - given the potentially adverse results which could follow, in the loss of national monetary control and the implications for the efficacy of national macroeconomic policy. Although municipal development financial institutions (MDFIs) are not new, according to Dillinger (1989), "the merits of this approach to local government finance remain unproven".

Borrowing - from central government or intermediaries - may be undertaken in order to finance major capital investments that cannot be paid for from today's recurrent revenues. This is generally done through the creation of a MDFI, which offers a diverse portfolio of financial issues, and has greater flexibility than local governments. These specialised institutions are said to have the advantage that the risk of local government default means that they are more likely to scrutinise creditworthiness more carefully, select better projects, and enforce repayment more rigorously. Municipal credit institutions have an established record in continental Europe and Japan (Dillinger, 1989). In developing countries, an MDFI usually begins with an initial injection of funds from a donor and a counterpart contribution, usually larger, from the central or state government. The World Bank has provided seed money to establish or stabilise a number of MDFIs. Where, as in Latin America, these financial institutions depend solely upon loans from central government to finance their lending operations, the resources mobilised would generally not enlarge the total public sector 'pie'.

In Brazil, the state governments of Santa Catarina and Parana have set up funds from which creditworthy municipalities can borrow for capital investment, provided that they have a satisfactory investment plan and are willing to accept technical assistance to help increase local revenue raising capacity (Rondinelli, 1990) In Senegal, local authorities can submit projects to the Comite National des Prets, for technical evaluation on the basis of established allocation criteria. The next stage is financial evaluation, and then financing by the Banque Nationale de l'Habitat (a central government financed institution originally created to administer low-income housing programmes). This credit system is reported to be yielding satisfactory results in terms of the number and size of operations undertaken, but an improved capability to assess the actual implementation of the projects and the use of funds is required (EDI 1989). Jordan's MDFI has been described as "notably successful" (WDR 1988). An autonomous Cities and Villages Development Bank, established by the government in 1979, provides investment finance and technical assistance to municipal and village councils. It has helped finance the expansion of physical and social infrastructure (such as roads, schools, clinics, water supply and electricity) to virtually all settlements, even the small and remote. It has also been "instrumental in maintaining and improving the quality of infrastructure investment through its standards, appraisal and monitoring", and become increasingly involved in financial and technical advice for local authorities (p.164). The Bank has introduced criteria which helps municipalities to better forecast their debt service capacity.

Central governments tend to strictly regulate and limit the ability of municipalities to borrow. This is not only due to concerns about financial solvency. The monetary implications of unrestricted borrowing might have significant inflationary repercussions, and undermine macroeconomic stabilisation policy at the national level.

In Nigeria, fear of excessive borrowing meant that only the largest, most financially viable municipal governments were allowed to take loans. In Brazil, the federal government sets specific limits on both the amount of debt and debt service that municipalities can incur (Rondinelli, 1990). In Zimbabwe, urban councils raise about 80-90% of recurrent revenue from ownsources - taxes, fees, licences, tariffs and rents. The capital budgets in the form of loans from the central government, or from the Local Authority Pension Fund. Only two councils, Harare and Bulawao, are allowed to borrow on the open market and can issue stocks and bonds (Moyo, in EDI 1989).

Finally it should be noted that the effective rates of interest faced by local governments will tend to vary, depending upon the nature of the bondholders, e.g. whether borrowing from the central government, or overseas banks or local residents. In some cases there have been problems with repayment. MDFIs in Honduras, Kenya and Morocco, for example, are reported to have experienced (and tolerated) substantial arrears (WDR 1988). Empirical evidence on total public resource mobilisation This is a crucial, but generally neglected dimension of decentralisation in developing countries. There are a number of empirically testable hypotheses concerning how the aggregate level of revenues generated might be influenced by the structure of government, in particular, when spending and taxation decisions are made on a decentralised basis. Yet there is very little in terms of rigorous quantitative evidence. Ideally, we would want some specific studies of developing countries - both time series (before-and-after comparisons), and cross-country data (between different countries with differing degrees of decentralisation). The most relevant strand of research to date has produced a lively debate (among some U.S. academics), over the relationship between decentralisation and government size. Brennan and Buchanan (1980) put forth the view, by analogy with the conventional theory of monopoly in the private sector, that a monolithic (i.e. centralised) government will systematically seek to maximise tax revenue. This led them to the proposition that the size of the public sector varies inversely with the extent of decentralisation. From the perspective of the above-mentioned authors, (which is obviously somewhat different than the present one), decentralisation is endorsed as a means of constraining total public resource mobilisation. Marlow (1989) explored the hypothesis that 'Total government intrusion into the economy should be smaller, cetirus paribus, the greater the extent to which taxes and expenditures are decentralised.' According to Marlow's analysis, the degree of fiscal decentralisation affects public finance through the following avenues: (i) Decentralisation increases competition in the public sector, leading to a relatively lower tax burden - i.e. given mobility of residents and capital in pursuit of fiscal gains,

the greater the numbers of alternative tax jurisdictions, the less likely that 'excessive' taxes will be levied; (ii) Conversely, centralisation restricts the ability of states to compete (as in (i)) since a growing national share of total government money weakens the relative significance of local governments; and (iii) Centralisation may generate a greater reliance on inflationary finance given that only the national government is able to print money. Unfortunately (for our purposes) Marlow's empirical analysis of government spending was limited to the USA. It covered the period 1946-1985, and used expenditure as a measure of government size (rather than taxation). He found that decentralisation affects the growth of government negatively (p.266), and that the shifting of government responsibilities from the Federal to the state and local levels will contribute to a slowing, or falling, of public sector size in the USA. This conclusion is controversial. Oates (1985) argued that decentralisation would tend to increase total public resource mobilisation. In his empirical investigations, Oates had found in 1972, for 42 countries, regressing tax revenue (size of public sector) on a fiscal centralisation ratio (central revenue as a proportion of the total), a strong and statistically significant negative association. Controlling for income resulted in a still negative result, although it was not statistically different from zero at the usual confidence levels. In 1985, Oates explored the question again for 43 countries (18 industrial and 25 developing). The results were that, for the entire sample, the rank correlation between the size of the public sector and the extent of centralisation is strongly and significantly negative (table 11). A relatively decentralised public sector is typically large. However this result is misleading since (as noted above), there is a dramatic difference between the average developing and industrial country. Developing countries have small, centralised public sectors, compared to industrial countries. Examination of the co-efficients for the two variables shows that there is no longer a significant relationship between them. Within each of the sub-samples, decentralisation has little relationship to the degree of decentralisation. Overall then, Oates' results lead to the null hypothesis that de/centralisation and the size of government have little to do with one another. Further mobilisation in developing countries over time is needed, before and after decentralisation, to gauge the impact thereof.
Occasional Paper 13 - DECENTRALISATION: A SURVEY OF LITERATURE FROM A HUMAN DEVELOPMENT PERSPECTIVE

4. Participation

Modes of Participation The Intensity and Quality of Participation

The Instrumental Value of Participation

5. The Impact Upon Economic Activity

Practice

4. Participation
Participation facilitates in the political realm the pre-eminence of human development and selfdetermination, both committed to the ideals of freedom and equality. (Borter, quoted in Frenkel 1986).

At the outset, the 1991 Human Development Report emphasises "Development by the people -through appropriate structures of decision making, people must participate fully in the planning and implementation of development strategies". Participation can be defined briefly, for present purposes, as 'taking part in the formulation, passage or implementation of public policies' (Parry 1972). It is necessary, however, to be more specific as to the nature of participation in a particular context, and identify who takes part in what, when and how, and with what results. This is the task of the present section. A distinction can be drawn between participation viewed in terms of active, broad-based local participation in political decision making, as an end in itself; and participation as a means to affect the appropriateness of decisions and enhance the impact of public expenditures upon efficiency, equity, private initiative and so on. These aspects can be referred to, respectively, as the developmental and instrumental values of participation (Parry 1972). Although this distinction is not always clear-cut, it does usefully reflect two approaches to thinking about participation. The former is clearly important, and much of the relevant literature is concerned with participation in the developmental sense. However our ultimate interest in the present survey lies in the instrumental aspect - the impact of local participation, be it broad based or not, upon decisions about the allocation of expenditures for human development. Even where local participation is characterised as being "subject to elite domination", it is still possible that there would be a positive impact upon decisions about human development. This ties into the discussion of comparative priorities below. These issues are addressed in turn, under a number of headings; the mode, the intensity and the quality of participation, and then the impact of decentralisation and participation upon the nature of decisions made. First, as to whether decentralisation improves the modes of participation - we see that decentralisation typically involves the creation of additional formal structures, ostensibly aiming to increase local participation. Second, the question as to who effectively makes decisions at the local level is addressed - it is found that national officials and/or local elites often tend to dominate the process. Finally, and most importantly from our present point of view, the impact of local participation upon decision making and priorities is explored. Even in the absence of broad based

participation, the actual decision makers may be responsive to interests other than their own, so that the quality of decision making is enhanced. Developmental - Does Decentralisation Enhance Political Participation?
"Like a well-established and well-loved music hall act, participation and local government are words often found in partnership" (Bulpitt 1972).

Participation is an important theme of many writings on decentralisation. The emphasis tends to be upon its political dimension, exploring the extent to which decentralisation effectively involves people in decisions about, inter alia, local development. Managing one's own affairs is held to be an integral element of human development, the objective being to maximise the opportunities for individuals to participate in the decisions which affect their lives. Decentralization and participation are often positively linked. The underlying belief is that local decision making structures will facilitate the genuine involvement of all the people, in issues of direct concern to their needs and interests. This in turn assumes both the desire of individuals to take part in local government, and the existence of appropriate forums which allow and encourage the free expression of their interests. Local government is sometimes presented as a form of direct democracy, or at least as essential to the workings of the modern democratic state. Yet there is no necessary correlation between broad based local participation and decentralisation. The relationship depends basically upon the social, economic and political setting, and the type of decentralisation enacted. In some cases, 'true' devolution and increased participation were clearly not part of the national government's agenda; the objective was no more than a limited degree of deconcentration. A number of authors (e.g. Conyers, 1984; McGinn and Street, 1986; Collins 1989; Cassasus 1990; Samoff, 1990) have emphasised that decentralisation should be seen basically in terms of rearrangements of power, and tend to conclude that such policies often result in deconcentration, and primarily strengthen central power. Ideally we would want to examine participation (or the lack thereof) on the basis of concrete evidence, such as the holding, fairness, and competitiveness of elections, voter turnout, lobbying of representatives, etc... In practice one must resort to more secondary sources of information. Modes of Participation The mode of taking part in local decision making will differ according to the opportunities - institutional and informal - available, the interests and political resources of the potential participants, and the attitudes prevalent in the society (Parry 1972). It is generally inaccurate to describe a political decision in terms of a once-and-for-all act; rather there are typically a number of stages, ranging from putting an item on the agenda, through technical advice, discussion of favoured proposals, to enforcement of the final decision. For example, local participants may be more effective in putting matters of

local concern on the agenda, and much less involved in the technical aspects of decision making. The focus here is upon the institutional arrangements which have been created under the auspices of decentralisation. In several of the cases, there has been a negative relationship between 'decentralisation' and participation, and a tightening of central, often authoritarian, control. These examples typically, although not invariably, fall at the deconcentration end of the spectrum.

Under a military dictatorship in Chile, the devolution of responsibilities for basic social services to the municipalities occurred simultaneously with closer central government control, with previously elected local officials, including mayors and community advisors, being subject to Presidential appointment (Winkler, 1989). Zambia's announced policy of "decentralisation in centralism" sought to create an "integrated local administration system". The center transferred power to the district level, but through Party officials and political appointees, both of whom are subject to tight central control. In a sense, this deconcentration broadened opportunities to participate in local decision making, but this was strictly channelled through, and constrained by, the (one) Party structure. It resulted in a single politico-administrative structure (District Council) in each locality, which effectively enhanced the "leading role" of the Party (Conyers, 1984; Chikulo, 1985; Nooi, 1987). The reforms effectively strengthened central control over a wide range of affairs previously regarded as local (Olowu 1989). The functions of the Zambian state at the local level include "official socialisation, by which an official ideology and appropriate values and attitudes are diffused, and by which government action is represented as being in the common interest; (and) participation, in politics and planning, in order to maintain the legitimacy of the state" (Rakodi, 1988 p. 32). In the Sudan, the Local Government Act (1971) sought to achieve "maximum possible participation of citizens in the administration of their local affairs and thus reduce centralisation" (Olowu 1989). At the same time, the Act made Provincial Councils (which are presided over by a political appointee) responsible for, inter alia, "political enlightenment and people's mobilisation ... in accordance with the objectives of the May Socialist Revolution". The new system was said to have, above all, aimed at encouraging the widest possible participation in government. The evidence as to the reality of public participation is unclear. Opportunities for female participation were greatly increased by a 25% membership quota in the councils. There was, however, general confusion as to what level of participation was appropriate - given the existence of six councils at the lowest level, two at the intermediate, and the Provincial Council at the apex. Further, the Provincial Commissioner - concurrently a personal appointee of the President, provincial secretary of the Party (SSU), Treasurer of the Council and head of the public service in the province - was able to veto decisions with which he disagreed. In general, particularly in the south, the system failed to operate as (ostensibly) intended (Norris, 1983).

Over the past two decades, the Kenyan government has frequently affirmed its belief in the concept of decentralisation, and the involvement of wananchi (the people) in identifying, designing, implementing and managing projects and programmes in the districts. However, critics have argued that during the same period, elected local government has been weakened, and that the 'nyayo' philosophy of the President requires a close linkage between local government and national policy, as the latter is determined by the central leadership (Wallis 1990).

An overview of country experience suggests that decentralisation is more likely to be thoroughgoing, in the sense that there is genuine devolution of authority and efforts to promote participation, under liberal democratic/ pluralist national regimes. Where the nation state is authoritarian/one-party, the mode of decentralisation tends to follow deconcentration, which effectively preserves central control. This applies equally to capitalist and socialist regimes, where small cliques have captured the power of the state. At the same time, however, the converse does not necessarily hold - democracy need not lead to decentralisation. The foregoing proposition is supported by the above-mentioned country cases, as well as the experience of particular countries over time.

In Peru, in the period since 1949, there have been shifts in the attitude of national governments toward decentralisation. The authoritarian administrations of Odrria, Velasco, and Morales Bermudez have consistently rejected devolution and political delegation, preferring the greater control of deconcentration and bureaucratic delegation. The democratic governments of Prado, Belaunde and Garcia emphasised political devolution of decision making authority - the latter attempted to implement a sweeping programme of general devolution to the regional and local levels (Scmidt, 1989) In the Philippines, after the imposition of martial law in 1972, a marked intensification of centralisation was observed (Ruland and Sajo 1988). Local governments were seen primarily as a device to consolidate the regimes hold on power. Local elections were suspended until 1972. Urban barangays were created as the basic administrative units of the country. While the barangays were purportedly created as a means to broaden citizen participation, they were led by regime-loyal, hand-picked (until 1982) officers, and studies found that in low income areas they primarily exerted functions of political mobilisation and control for the regime (ibid p.27). In Liberia the militarisation of local administration followed the 1980 coup. Even after the promulgation of the new constitution and the ostensible return to civilian rule, the role of the military in local administration remained significant. In 1987, six of the thirteen county superintendents were military men in active service (Sawyer 1990).

The Intensity and Quality of Participation : Who Makes Local Decisions

Who participates and with what frequency? The intensity of participation refers to the proportion of the population that takes part in political activity (Parry 1972). A related question lies in the quality of participation, or its effectiveness - whether the policy outcomes are those intended by the participants and the direct result of their actions. Voting, for example, is sometimes regarded as qualitatively inferior because it is distant from the legislative outcome, and the average voter takes no further part in the decision making process, abdicating her will to the electoral representative. In any case it is doubtful whether voting in an election where there is only one party amounts to 'real' participation. Even where an electoral system is formally open, the outcome may be as predictable as in one-party states because the opposition lacks the necessary resources (wealth, skill and organisation) for success. Alternatively the policies and platforms of the 'opposing' parties may be so similar that there is not, in reality, any real degree of choice exercised in the act of voting. On the other hand, going to the polls may be the most which can be expected of the average participant, in developing and industrial countries. It has been argued that decentralisation, by multiplying the available modes of participation and by establishing institutional arrangements by which votes will be more closely linked with political consequences, improves the effectiveness of participation. Local knowledge of, and a commitment to, the community is said promote a desire to participate in local government. However there are numerous examples of formal attempts at decentralisation which have failed to engage the effective participation of local people in decisions about local development. Sometimes this is apparently because the national government never actually intended such a result to follow (as in the examples above). In other cases it may be in spite of the creation of structures ostensibly designed to facilitate local participation, often because of the dominance of decision making by official bureaucrats. In the following paragraphs, some of these are recounted. National Officials This survey has emphasised the likelihood that local government will be regarded simply as a technical and administrative arrangement within the national government machinery, which relieves the central authorities of direct responsibility of providing certain public services. The civil servants who staff the various central field agencies at the local level play a number of roles, including the provision of technical advice, the administration and co-ordination of services, and regulatory functions. It has been observed that they may also dominate local decision making, for reasons ranging from superior expertise to ulterior motives on the part of central authorities.

In Zimbabwe, channels for popular participation were created in the form of Village and Ward Development Committees, and District Councils. In practice however, it has been found that elected representatives were frequently overruled by the technically more competent central government officials, who have more direct control over the allocation of public expenditure (Mutizwa-Mangiza, 1990). In Senegal, administrative decentralisation reforms in 1972 sought to encourage popular participation in the management of local affairs. Through the active role

of local people in the making of development policy, the government hoped to counter the "rising tide of the malaise paysan" wherein the peasants were refusing to repay government loans or otherwise co-operate with government programmes in rural areas (Vengroff and Johnston, 1987). To ensure 'responsible participation', the affairs of each rural community (group of villages) were to be managed by a representative body, the rural council. Two-thirds of the councillors are elected (on a winner-take-all party basis), and one-third chosen by the general assembly of co-operatives in the rural community. The councillors (overwhelmingly male and poorly educated), often felt that their views were not taken into account in the initiation and execution of local development projects. The above-mentioned study concluded that the national administrative authorities were put in a position which allowed them to "exercise authority in a manner inconsistent with the spirit of the administrative reform." Moreover, the system adopted did "not represent a major departure from French administrative practice, in that the state and its agents retain full supervisory control over all aspects of local level actions". In Chinese cities, the Residents' Committees and Street Offices, the political organs which form the linkage between the people and the government, are supposed to provide opportunities for participation in the governmental process. However it has been argued that the Chinese government has a "top-down" approach toward community development which seeks to maximise popular support for national policy goals and Party officials (Leung, 1990). Seen in this light, the community structure serves as an instrument of the state to secure ideological unity. The cadres perform apparently conflicting roles - as representatives of the interests of the people, and government officials implementing government policies. The government, however, does not perceive this as a problem. Critics argue that the community structure of committees does not encourage the articulation of interests which diverge from the official mass line. And correspondingly, some surveys have found that half the people saw the resident committees as 'nasty and meddling', another quarter viewed them neutrally, and only a quarter of the sample regarded them as helpful (ibid p. 204). The importance of the structure as a mechanism of political control was allegedly demonstrated in the aftermath of Tiennamen Square in 1989. The street offices and residents' committees were immediately mobilised to support the government, through organising political study groups, putting up pro-government banners and assisting the police in tracking down dissidents (Leung, 1990).

Another, perhaps extreme, example is the set of decentralisation proposals developed by the South African government, to accompany the amended constitution of 1983. The state's guidelines for black community authorities purported to envisage maximum devolution of power to the local level, and minimum administrative control. Contrary to the stated objectives however, decentralisation has been interpreted as a strategy for extending central control, by depoliticising the population, by controlling the actions of local government institutions and by using the local authorities as a "training ground for

collaborative political leadership" (Samoff, 1990, pp.522-23). In the short term, it was also intended to deflect popular protests from the national to the local government, and thereby reinforce central authority (id.). In some cases, the impact of decentralisation upon the intensity and quality of local participation is controversial. Tanzania is one such case. The reforms embarked upon in 1972 included the creation of deliberative assemblies at all levels, which involved both elected representatives and the relevant government functionaries. Maro (1990) concluded that this had thereby increased the level of popular participation in the drawing of village plans, through the village council and committees. Slater (1989) however, regards the country as an example where the "rationalisation and consolidation of centralised authority lay at the roots of the spatial restructuring of state power, so that decentralisation was more illusion or myth than hard institutional reality" (p. 514). That author argues that there was little real attempt to facilitate mass participation in the decision making structures, and little dialogue between government staff and the people to find out about actual problems, potentials and needs. Final decisions about regional and district plans and personnel all rested firmly with the center (Ministry of Economic Affairs and Planning and the Prime Minister's Office). The machinery for popular participation provided in each Ujaama village was under the general supervision of the centrally appointed village manager. According to Kasfir (1983) the degree of participation in the newly created institutions was actually less than that prior to the reforms. Similarly, Samoff (1989) concludes that in practice, Tanzania experienced deconcentration, which largely reinforced central authority and converted village councils into development advisory committees, dominated by administrators and technicians. Local Elites It is sometimes argued that community attachment and knowledge will lead to increased rates of participation in local affairs. But even where there is a substantial measure of devolution, participation on the part of those people previously excluded from decision making structures need not follow. Empirical studies have emphasised the importance of high social status (income, occupation and education), characteristics not necessarily confined to those with a high degree of community attachment, as indicators of high political participation rates in local government. The studies available suggest that power structures in small communities are often oligarchic in character, and that political activity is typically the preserve of the few and the more important the decisions are, the fewer the number of people involved (Bulpitt 1972). This contradicts the developmental objective of broad-based participation. Popular election does not always circumvent this problem. Given the distribution of assets (land), income, education, and client-patron relations, local elections tend to favour the rich and "well-born", especially in highly stratified societies. The phenomenon of elite domination is not unique to local governments, nor to developing countries. The role of elites is one of the basic problems for direct, as well as indirect (representative) democracy. According to the U.S. political scientist Ostrom, the existence of an elite is unavoidable for any political system, be it democratic or not. It

might be added that the postulate of participation in the sense of an equal influence of everyone is unrealistic inasmuch as it presupposes everybody having such an interest, in the face of a multitude of 'private' endeavours which might be much more effective, at least from the individual's point of view (e.g. meeting basic consumption needs). Although we do not know how governmental structures affect political attitudes and behaviour, it seems reasonable to believe that the characteristics of local representative democracy will resemble, to a certain extent, democracy at the national level (Bulpitt 1972). Local systems of organisation are power systems, quite as much as national power structures. It has been argued that new local organisations are quickly taken over by those holding control in the existing structures (Ralston et al 1983).

A study of Iloilo City in the Philippines found that local politics was marked by a high degree of competitiveness, especially prior to martial law, when voter turnout averaged nearly 81% between 1947 and 1971. It had fallen markedly by 1980 (to 74%) reflecting a one-sided contest under the auspices of authoritarianism. Yet political participation was not broad based, in the sense that the upper-middle and upper classes dominated decision making, and competed for elected positions. In one sample of local politicians, 30% were lawyers, 26% other professionals, 29% businessmen - there were no lower class representatives on the council. "Wealth, personal connections, family prestige and a wide clientele are still the major criteria for political success" (Ruland and Sajo 1988). Even within democratic political cultures in the North, the incidence of participation varies significantly, as does the kind of people who seek to participate. In the U.S. only 15% of those aged between 18-24 voted in the last Congressional election. The better educated and better off are far more likely to participate in politics and government than the uneducated poor. [In some industrial countries (eg Australia) local government elections are notorious for low intensity of participation (in terms of turnout at elections).]

In many developing countries, women have been excluded from traditional and modern forums of decision making, at all levels of government. The extension of participation requires both the erosion of male privileges and prejudices, and the creation of adequate educational, training and other facilities (such as family planning). It appears that such comprehensive changes require, at the very least, positive governmental action at the national level.

A field study in Nusa Tenggara Timor, Indonesia demonstrated the difficulties in eliciting the participation of women in development projects. Each village in Indonesia is headed by an elected village chief and his assistants. Whilst issues are discussed in a musyawarah (community meeting), decisions are effectively made by the chief. Women lead relatively isolated lives and rarely attend musyawarah. The main vehicle to stimulate the involvement of women is the Pembinaan Kesejahteraan Keluarga which, at the village level, is headed by the chief's wife (in accordance with cultural traditions). In the villages studied, this body was functioning poorly and the chief was actually making the decisions.

Some argue that separate forums for women should be established, rather than included in mainstream organisations (Yudelman 1986). The reasons against integration are that women want to run their own organisations, and to earn and control their own income (in the case of income-generating projects). On the other hand, it is stressed that separate is not equal, because women tend to be politically and economically marginalised by such efforts. Quotas are one possible means by which the observed tendency for elite domination can be redressed. They can ensure a minimum representation of deprived and minority groups.

In the Indian state of Karnakata, the composition of the mandal panchayat (group of villages) encourages participation by groups commonly excluded from local decision making - 25% of the seats are reserved for women, and a minimum of 18% of seats are reserved for members of scheduled castes. Some cynical commentators attribute this to the Janata Party's desire to incorporate marginalised voters into its patronage network, in order to challenge "the Congress machine" (Slater and Watson, 1989). In Vietnam, adequate representation of poor peasants in the decision making bodies of production co-operatives was sought through reserving two-thirds of the seats on the relevant committees for low income groups (Majeres 1985).

In the USA since the 1960s some federal social laws have provided that the programmes have to be administered with "maximum feasible participation" of the local poor. These national programmes were administered through a (centrally designed) ad hoc local forum which ignored the traditional local structure in favour of the underprivileged (African Americans, Latinos etc.). The reviews of this approach have been controversial. But there is agreement that the scheme influenced local community power structures to a much greater extent than had been anticipated by the national politicians who introduced the provision (Frenkel 1986). The Instrumental Value of Participation - The Impact of Decentralisation on Local Decision Making Instrumental theories portray participation as a means by which individuals can ensure that their interests are defended and promoted, and as a way to facilitate better information and communication, and therefore more effective government. Participation is also expected to induce enterprise, initiative and imagination which will benefit the whole community. Unfortunately the relationship between decentralisation and the degree of participation facilitated thereunder on the one hand, and the impact upon other objectives of human development on the other, is not clear-cut. There is mixed evidence as to the instrumental aspects of participation in a decentralised system. McGinn and Street (1986) conclude their discussion of (educational) decentralisation in the following terms: "non-democratic governments will not use decentralisation to broaden democratic participation (no matter what their rhetoric), and

whatever benefits result from increased efficiency will be distributed inequitably in an inequitable society" (p.490, italics added). Yet we will see evidence which contradicts this thesis. Indonesia, under a manifestly authoritarian regime, has achieved through the substantial decentralisation of expenditures on basic social and economic infrastructure, impressive strides in human development over the corresponding period. Such advances were equitably distributed across the archipelago, especially so in favour of the impoverished eastern islands. And in Taiwan, measures of decentralisation adopted in a state which was far from democratic have contributed to relatively equitable patterns of growth and infrastructure development. It has already been observed that even where there is a thoroughgoing devolution of decision making power to local government, the poor may still be effectively excluded. Beyond simply labelling situations as being "subject to elite domination" however, it is necessary to pursue the issue further. Are local spokespersons, albeit members of the local elite, more appropriate and/or beneficial than the alternative of central control? Elite groups may act solely out of self interest, or be regarded as more capable representatives of local communities. Similarly, where civil servants and nominated individuals dominate local decision making they are not, automatically, less responsive to local demands. Local constituents may prefer to contact field agency officials rather than their locally elected representatives. Finally it is noted that the fact that most important decisions are made in the national capital does not automatically mean that local interests and demands will automatically be ignored or slighted. These are largely empirical questions. It is necessary to look at the priorities and behaviour of those who occupy local positions of power, and compare this to the revealed preferences of central authorities in the capital. In some cases, the outcomes will be better, in others worse, and sometimes perhaps no different than that which would have resulted under a centralised system. The available studies suggest that power structures in small communities are conservative in their policy attitudes (Bulpitt 1972). The notion of 'community' which often underlies discussions of local government has been criticised; community harmony may be the result of repression and community integration the result of conflict. It has been argued that the dominant value of small community life and local government is fraternity, which emphasises consensus and stability rather than conflict and change (Wood, cited in Bulpitt 1972), to the disadvantage of the poorer members of the community. The following paragraphs reflect instances where the decisions of local government have favoured the better-off groups in the community. In Nepal, where political parties were (until recently) banned, it was reported that the people which traditionally held social and economic power (such as wealthy farmers) have used the system instituted under the Decentralisation Act (1982) to benefit themselves (Bienen 1990). In Bangladesh, the chairmen and elected members of the union parishad (the local government unit in closest contact with the people) are drawn from the rural elite, because of the influence of traditional patronage relationships. It has been argued that an

authority that formerly entailed some reciprocity has been invested with official sanction and partisan support, thereby facilitating the ability of the local elite to further its own interests at the expense of the community. Powers have been transferred to members of the local elite, and to centrally controlled and recruited civil servants (Khan 1987). As noted above, local government in Iloilo City in the Philippines has been dominated by the upper classes (Ruland and Sajo 1988). This was said to have major repercussions on urban development processes and priorities - "basic socio-economic issues confronting the poor figured lowly in the value system of the city's decision makers" (ibid. p. 273).. In a country where unemployment was approaching 20%, this problem was only ranked 8th among 17 major problems confronting the city, and social inequity ranked 12th. It has been observed in East Africa that the main beneficiaries of community development schemes have been relatively privileged social groups - those with the initial resources, skills or influence to help ensure that 'self-help' schemes reflect their own particular needs. Some argue that centrally-sponsored institutions are inherently likely to 'co-opt' popular action. This implies that local decision making will not alter the distribution of the benefits of public expenditure, insofar as official decision making structures operate so as to modify perceptions about local needs.

In Israel, a highly centralised and politicised nation, Matnassim were established in order to provide an array of social services, and to facilitate citizen participation, through the majority representation of local residents on the Governing Boards. In practice however, local representatives tended to identify with a centralistic, homogenising view of society, more than with local interests. This was attributed partly to the fact that the political careers of local politicians are tied into national political parties. One study concluded that local Matnas have not been able to independently develop programmes of local interest - and that it was "not effective as a vehicle for community-specific development within the heterogeneous Israeli society" (Havassy and Yany 1990).

One drawback of central field agency domination, from an instrumental point of view, is that national officials may tend to lack the time and resources to develop adequate contacts with the public. It has been said that these officials are likely to have an "enforced neutrality", avoiding contacts and initiatives wherever possible (Bulpitt 1972). Obviously the foregoing evidence about the impact of local decision making forums upon actual distribution of the benefits of public expenditures is rather scattered and anecdotal. More comprehensive evidence and analysis is presented below in Part IV, under the discussion of the comparative priorities of local and central government. 5. The Impact Upon Economic Activity The expansion of income earning opportunities is a crucial dimension of human development. Yet the impact of governmental structure upon economic activity is

difficult to pinpoint, especially given the multitude of associated variables such as macro economic policy, development strategies and ideology. Ideally we would like to be specific, and obtain evidence as to whether the decentralised provision of economic infrastructure has actually promoted local economic activity. Unfortunately there is little direct evidence either way. In principle, decentralisation may promote local economic activity through several means, including an increased infusion of capital and other resources, the more extensive provision of infrastructure, and a more effective 'enabling environment' than would have been the case under a centralised system. In this light it has been claimed that countries with effective local government systems have been much more dynamic and successful economically than those under centralised control. It is further argued that there exists a causal correlation between decentralisation and increased economic activity (Allen 1987). Indeed the general assumption in much of the literature on decentralisation appears to be that decentralisation promotes economic development (Olowu, 1987; Olowu and Wunsch 1990). According to Tocqueville, in his criticisms of the nineteenth century European administrative system, the necessary creativity and popular enthusiasm to sustain the development process might be lacking in the absence of effective basic local government systems (Tocqueville, Democracy in America, cited in Oluwu, 1987). As suggested above, the potentially positive implications of decentralisation for private sector activity are various. First, where decentralisation is accompanied by increased public resource availability and capital inflows, this will have direct and secondary stimulative multiplier effects. For example, the construction and maintenance of local infrastructure, such as roads, water supply and electricity, is likely to involve local contractors and wage labour. Second, as we saw above (p.*) in the discussion of economic infrastructure, decentralisation is likely to lead to more appropriate selection of the type and location of facilities, and better long term maintenance. Thus decentralisation can contribute to improved physical conditions (transportation, energy sources, communications etc.) which reduces rural isolation and improves the profitability of local enterprises, and thereby stimulates local economic activity. Third, local governments may be more supportive of "endogenous entrepreneurial development". The latter is defined by the OECD (1986) as development built on local assets, expanding existing enterprises and creating new ones, which is initiated and managed locally. Governments at the local level may be better able to formulate policies which are appropriate to local conditions. Entrepreneurial energies may be unleashed from the burden of central bureaucratic regulation (assuming, of course, that these are not replaced by local red-tape). Thus it has been observed that "because they need support that is individualised to local conditions, entrepreneurial initiatives appear to survive and flourish best in settings that permit flexible and localised public management (OECD 1986,p.33). Finally less immediate benefits might also be forecast, such as those flowing from a labour force whose skills better meet the needs of local employers. Devolution of decision making power to the local level may also benefit local entrepreneurship by helping to overcome structural constraints. The rural economic environment is

characterised by large distances from the urban centres where both public and private decision making tends to be conducted and financial resources and management information concentrated. Dispersed rural enterprises have tended to be relatively isolated and have poorer knowledge of the economic activities in their immediate environment and beyond. Local governments may be well positioned to formulate and implement policies which are supportive of local business including such services as management assistance and market information. It is said that the physical and social distances which separate central officials from rural entrepreneurs, and that the indiscriminate character of regional policies have inhibited the development of fruitful policies by the national administration (OECD 1986). Caputo (1988) is enthusiastic about the potential for local governments in "enabling local economic growth". This is to follow from their ability to adopt a management approach "based on the active participation and promotion of many small businesses" (p.111). Access to public contracts - through local government calls for tender for the construction and equipment of decentralised public facilities - is presented as a significant factor in the dynamic growth of small businesses. The other "main lines of action" for local government lie in the establishment of systems to gather and analyse information on the dynamism and constraints of local small businesses, support for small business associations, and involvement in the creation of effective permanent channels for consultation and participation (p.115). The author argues that programmes for supporting small businesses are best promoted at the local level. At the same time the influence of national macroeconomic policies must be borne in mind - where central economic managers are seeking to stabilise or contract the economy, any local stimulatory measures may well be overwhelmed in the larger context. Moreover national development policies - industrial, agricultural, trade etc. - will generally have a substantial impact. A national system of direct communications, highways and railways is also important, especially for local 'export' industries. Practice There are a number of cases where decentralisation has stimulated local economic activity. Sometimes this has followed the construction of economic infrastructure; in other places it has been attributed to the capital injected through transfers to impoverished localities under a decentralised system.

In Taiwan, the agricultural land reform programme had a significant effect on population distribution in the 1950s and 1960s. It created some 200 000 new owner-farmers and greatly increased job opportunities in rural areas. From the early 1970s the government began to focus on spatial development issues. The location of public utilities has been fairly well spread throughout much of the island. The road network is evenly developed over the western and eastern coastal plains of the island and there is a good rail system. This facilitates the dispersed distribution of industrial establishments. The perceptible growth of population and industry in the central and eastern regions, and the less rapid growth of the

Kaohsiung and Taipei metropolises in recent years suggests that the decentralisation of infrastructural improvements across the island has helped to avoid over-concentration of population and economic activities in the metropolitan areas (Tsai 1990). In Indonesia, the major redistributive schemes and central transfers adopted under the auspices of "decentralisation" are said to have been a significant factor in explaining the healthy growth rates of relatively poor regions, especially in the Eastern islands (Hill 1991). Decentralization in Tanzania sought to promote small-scale industries in a selfreliant fashion (Maro, 1990). The reforms increased the use of local resources, raw materials and skills. There was also evidence that these industries produced goods and services which could otherwise only have been obtained at the expense of high and foreign exchange costs. Positive economic consequences do not flow automatically however. There are instances where decentralised governmental structures have not provided a positive link promoting local economic activity. Sometimes this can be attributed to an allocation of government responsibilities between different levels of government in a manner which limits any positive local impact; in other cases, there are continuing inadequacies in infrastructure at the local level.

Nigeria's 1976 reform created a three tier structure of government - federal, state and local - the third being based on local units which varied greatly in size and population. No distinction was made between urban and rural areas. The range of functions allocated to local government councils included primary education, health services and roads, but did not extend to water supply, small scale industries and co-operatives. Several negative consequences are seen to have resulted from these arrangements, such as lop-sided infrastructural landscape in favour of local government headquarters in urban centres and the absence of agricultural programmes in most areas (Apkan 1990). In the Philippines the development of urban hinterland centres was pursued in order to rectify the country's grave regional economic imbalances. The key strategy, known as the Regional Cities Development Project (RCDP), involved infrastructural development, strengthening management capabilities of the city government, and "the creation of conditions conducive to the expansion of an urban economy". Ruland and Sajo's (1988) study of the impact of the scheme in Iloilo City found that some of the infrastructural components - such as a modern port complex and a well-developed road system - improved the foundation upon which to build a sound manufacturing sector. Yet serious infrastructural limitations remained in the inadequate water system (whose supply had not increased in the past 20 years) and electricity which was expensive and insufficient to supply heavy industry. The study concluded that overall the RCDP had not yet induced economic growth in the city. On the contrary the prevailing economic crisis left the infrastructural improvements to fall into "an economic vacuum".

There has been a considerable research and debate on the nature of industrial decentralisation, and the role of the government therein (see e.g. Bell 1987; Wellings 1988; World Bank Spatial Policies Programme). Decentralisation here refers to the relocation of industry from the metropolitan areas to the periphery. The dominant opinion is that it is primarily a result of government intervention in the provision of incentives and development of infrastructure. An alternative view is that industrial decentralisation is largely a spontaneous response to market forces.

This issue has been discussed extensively in the South African context, where a "vast quantity of money has been paid towards incentives and the development of infrastructure in far-flung industrial decentralisation points (IDP)"(Wellings 1988). The IDPs were able to offer infrastructure and services to industrialists. Surveys of local companies have shed some light on the factors bearing upon firm location decisions, revealing that the incentives were often the major attraction of the IDPs and that many firms would relocate if the subsidies were removed. On the other hand it has been argued that although incentives may have been decisive in some cases, the scale was substantially in excess of what was required, and that the process of relocation would have continued in any case (Bell 1988). The areas were also attractive to unskilled labour intensive industries, since labour was relatively cheap and unorganised. [As far as I can tell the controversy is unresolved.] World Bank research (published in Rogerson 1986) has suggested that Sao Paolo has been undergoing a 'polarisation reversal', so that national patterns of growth and development of increasing urban concentration have ceased and the processes of spatial decentralisation have begun. This "slow rearrangement of economic activities in space" is said to have occurred in response to market signals, and in the absence of strong policy interventions. Industrial surveys found that decentralised development depended primarily upon local initiatives and the emergence of new industries, rather than relocation or establishing new branches of existing plants. The underlying factors were diseconomies of scale (congestion, pollution, high labour and land costs); mandatory licensing for all new industrial investments in heavily polluted areas; and, most importantly, the improved availability and quality of industrial infrastructure in the regions. The study concluded that "instead of fostering often expensive large-scale industrial parks designed to lure metropolitan-based establishments, it appears more valid to channel efforts into stimulating local enterprise initiatives" (Rogerson 1986).

In sum, it appears to be generlly agreed that the local development of infrastructure, on a comprehensive basis, is likely to have a positive impact upon local entrepreneurial activity employment and economic growth.
Occasional Paper 13 - DECENTRALISATION: A SURVEY OF LITERATURE FROM A HUMAN DEVELOPMENT PERSPECTIVE

6. Equity

The Extent and Impact of Central Grants The Expenditure Effect The Taxation Effect

7. Obstacles

Political constraints Bureaucratic obstacles The role of NGOs and traditional structures

6. Equity Measured in terms of various indicia of human development, there are marked disparities between rural and urban areas, regions, and income groups in developing countries. This has been an important theme of the HDRs. Many examples can be cited (see HDR 1990 pp 29-33; 1991 pp.26-7). Thus the distribution of central grants, public expenditures, and the burden of taxation under a decentralised system are all critical aspects of the present survey. In this section the theoretical arguments and any available empirical evidence are brought to bear upon these issues. It is important to recognise the multi-dimensional nature of equity. In the present context, one must distinguish between inter - district (between different parts of the country), and intra - district (the position of different socio-economic and gender groups within the locality vis- -vis each other) aspects of equity. Improvements or deterioration in these various elements of equity need not coincide. Indeed it appears that while decentralisation can promote equity among different groups within a region, through increased local public expenditure and the wider provision of public goods and services, there is a risk that disparities between districts may worsen due to, for example, different tax bases and inadequate fiscal equalisation. The former tendency points in the direction of greater decentralisation, whilst the latter suggests that the central government should at least retain a strong redistributive role. There are several critical factors which determine how de/centralisation influences equity. These include central grants, expenditure and taxation effects. It is clear that no uniform generalisations may be forthcoming about the impact of decentralisation on aggregate inequality, since the impact of the various factors may be working in different directions. On balance, the evidence uncovered by the present survey suggests that aggregate inequality is likely to be predominated by worsening inter-regional equity, largely because of the absence of effective central redistributive policies. Intra-district

If decentralisation leads to increased local public expenditure and expands the provision of public goods and services, a narrowing of intra-district disparities in access among different socio-economic groups would be expected to follow. Moreover, if decentralisation of decision making is accompanied by increased participation of representatives from a wider variety of political, ethnic and social groups, greater equity in the allocation of public resources is presumed to be more likely (Uphoff and Esman 1974). Whether the latter is likely to eventuate was discussed in section 3 above. The reader will recall that many commentators are pessimistic as to the likelihood of active broad based participation in decision making under decentralisation. According to Hughes (1985), "too frequently the local beneficiaries are not the poor, as much as elite groups and political factions controlling the local authorities" (p.59). On the other hand the empirical evidence brought together on the topic of comparative priorities in the present survey suggests that local governments tend to spend a higher proportion if their budgets on social services, and on more basic levels of provision therein, than do central governments (see Part IV). From a theoretical perspective, inequalities within specific localities could worsen under decentralisation, because the redistributive functions of local government are thereby limited for two related reasons (Oates, 1985): (a) On the taxation side, "sorting out along Tiebout lines" will imply relatively little scope for redistribution from the wealthy to the poor within jurisdictions; and (b) Local redistributive expenditures will be constrained by the fear of attracting the mobile poor with relatively generous local support programmes. This suggests that the actual scope for assistance to the poor within the locality will be more circumscribed under a relatively decentralised fiscal system. Inter - district The impact of decentralisation upon spatial disparities in human development is potentially two-fold. Decentralisation, even in terms of its more limited mode (deconcentration), can reduce spatial inequalities in access to basic education and health care, clean water, decent income earning opportunities and so on, through the wider provision of public goods and services. On the other hand however, if decentralisation extends to the mobilisation of public resources (i.e. revenues as well as expenditures), existing disparities are likely to be exacerbated. This follows from the observation that the poorer regions of the country will have correspondingly lower fiscal potentials. In practice, there is evidence which reflects the operation of both scenarios. A number of studies support the proposition that inter-district disparities may be reduced through the wider provision of public goods and services. Rondinelli et al (1983) list among the positive results of decentralisation that "the access of people living in previously neglected rural regions and local communities to central government resources

and institutions has increased, if only incrementally, in most developing countries that have decentralised" (p.50).

In Morocco, the Local Authorities Development Fund (created as part of the decentralisation effort) distributed funds throughout the country on a far more equitable basis than its predecessor, the Communal Supply Fund, which had mainly been drawn upon by richer, urban communes (Cheema 1983). Maro (1990) found that in Tanzania, data on clean water, health facilities, primary education and roads showed that decentralisation had led to: (i) increased government expenditure on these services, which resulted in increased provision throughout the rural areas, and (ii) a drastic reduction in user-distance to basic social services, tending toward uniform access to services in all districts, and therefore a reduction in inter-district disparities.

The Extent and Impact of Central Grants The center has a potentially powerful redistributive role, in collecting revenue on the basis of ability to pay, and allocating funds on the basis of need. The latter can be done either through a bias towards poorer areas (tackling inter-district inequality), or via specific grants which tend to benefit the less well off (e.g. for primary education), thereby enhancing intra-district equity. The available evidence relates mainly to the impact of grants upon inter-district equality. It has been widely argued by writers on fiscal federalism (e.g. Musgrave 1959; Oates 1974) that the national government has the primary responsibility for initiating and coordinating policies so as to achieve horizontal equity - i.e. that policies for redistribution are most appropriately set and implemented centrally. This is partly because there should be a single level of government which has the role of determining the extent of redistribution, lest different decisions made by various levels of government undo the redistributive actions of the other. Moreover, the marked spatial variations in resource availability point to the need for the center to adopt the predominant role in redistribution (as in, for example, Indonesia - see case study). There is also the 'Tiebout effect' of population mobility, which limits the feasible extent of variance between independent redistributive measures at the local level (Helm and Smith 1987), and suggests the need for a strong central role in promoting an equitable distribution of public resources. (It is noted that the need for redistributive grants conflicts, to some extent, with the presumed need for a substantial degree of financial autonomy, discussed in section 2 above.) The central government may undertake substantive schemes of redistribution through grants which favour local governments in poorer parts of the country. In practice however, it appears that the prevailing fiscal arrangements under decentralisation - by design or otherwise - generally fail to enhance inter-district equity.

In Chile, the system of central grants was altered from one which redistributed in favour of poorer regions to one which made allocations on a per capita basis. Inadequate central financing failed to cover the cost of service provision (as it was supposed to), so that poorer municipalities could only afford inferior services thereby worsening existing inequalities between regions. Zimbabwe proceeded on similar lines to Chile, in that many recurrent central grants are allocated on a per capita basis with no consideration of community fiscal capacity (Winkler 1989). In India, it has been said that federal education grants are made primarily on political grounds and serve only to exacerbate spending disparities (Tilak, 1989). In Bangladesh, the fiscal system transfers resources from the urban to the rural sector (where 85% of the population lives). At the same time however, development fund grants are allocated on a per capita basis without regard to need, and the infrastructure grants appear to be distributed on an ad hoc basis (Bahl, in Schroeder, 1989). In Botswana, where the District Councils depend upon central transfers for over 80% of total recurrent expenditure, the system of resource allocation is described as "haphazard", resulting in the "uneven provision of local government services in different parts of Botswana" (Tordoff 1988 p.197). By way of contrast there is some evidence which reflects attempts by national governments to alleviate the inter-district equity problem, through grants which offset differentials in local tax bases.

In Uttar Pradesh, grants are distributed to the districts on a criterion of "backwardness" - taking total population, number of scheduled castes (who tend to be landless) and development of infrastructure (roads and electrification) in the district. Under this arrangement, which was introduced with decentralisation, the backward districts get more than in the past, and inter-district disparities have tended to be reduced (Sanwal, 1987). In Nigeria, constitutional changes over time have led to an equalising trend in federal transfers. Until 1978 transfers were made on the basis of derivation, as well as other criteria which focused more upon redistribution. In successive constitutional amendments, the latter gained in importance. The heavily populated and poorer northern states have been able to reduce the influence of the derivation principle in transfers and extend, first, the population criterion and, more recently, the "additional need" factor. Thus, new formulas have further increased the importance of the fiscal equalisation principle (Hinchcliffe 1989 p.445). In Brazil the 1967 tax reform led to the redistribution of public funds from richer to poorer regions (Graham 1987). The basic formula used by the federal government weighted the amount received according to the level of development. For every 100 cruzeiros collected by the federal government in taxes set aside by law for transfer to the states, the North received back 210 cruzeiros in 1976; the states of the South and Southeast regions receive less than 25 cruzeiros for each 100 they collect.

In the education sector, the World Bank found that the formula for distributing federal education moneys to the states in Brazil is highly redistributive. Similarly, Eastern Nigeria and Kenya have taken actions to redistribute government grants in favour of poorer community schools (Winkler, 1989).

The Expenditure Effect Decentralisation may enhance equity through increased public expenditure on those areas of most benefit to the poor. Improved provision and utilisation of basic social and economic services can follow. This relates to the discussion of relative priorities in Part IV below. In the following sections, the expenditure effect of decentralisation is appraised from the perspective of equity, both intra and inter-district, in the specific contexts of health and education. Different characteristics and concerns tend to be relevant to the each sector. In practice it is found that decentralisation does tend to expand the provision of basic social and economic infrastructure. Yet inequalities persist, and are sometimes exacerbated. This may be due to disparities in resource allocation or in utilisation patterns, or in the ability to respond to incentives under a decentralised system. It is a moot point as to whether this leads to a more or less equitable situation overall than that which might have existed under a centralised system. Health care Equity is an important consideration in the health context, where access to services (and indicators) varies markedly across different regions, income groups and sexes (HDR table 9). By increasing access throughout the country, decentralisation may help to reduce these disparities. However the limited evidence available on this point tends to suggest that the expansion of health resources under a decentralised regime does not tend to be equitably distributed, even within regions.

The most positive example of an equitable expansion of services across regions under a deconcentrated system is provided by Tanzania. Maro (1990) found that there was a clear bias in favour of regions that had fewer facilities prior to decentralisation, such as Dodoma, Coast and Mwanza. The result was a dramatic increase in the provision of health facilities. In 1972, only one in eight people lived within five kilometres of a hospital; in 1980, 60% of the villages either had a health facility or were within five kilometres of one. In Indonesia, between 1969 and 1983, in tandem with increasing decentralisation of public expenditures, there was a rapid expansion of health services. More recently the Posyandu (community-based integrated health post, mostly concerned with child survival) has been extended to all parts of the country. Currently, the Kecamatan (sub-district) health center provides the administrative framework within which most basic health care activities take place.

Yet despite this substantial deconcentration of health operations, there exist large disparities in utilisation patterns among different groups in Indonesian society which impact adversely upon intra-district equity. It has been found that, on average, the insured population (civil servants and their families) used the health services four times more frequently than the rest of the population. In one provincial study, the top income decile made up one-third of all hospital inpatients, one-half of hospital outpatients, and one-quarter of all health center visitors (Gish et al 1988).

A recent critique of the decentralisation of health services in Mexico provides an interesting counterfactual-type study (Gonzalez-Block, 1989). It also introduced the important variable of macro-economic conditions. The study covered two states in the poorer parts of the country, one of which decentralised to a significant extent (Guerrero), whilst the other did not (Oaxaca). It found that intermunicipal disparities worsened in decentralised Guerro. The statistics used to indicate health need satisfaction were: the number of general consultations; extension of service coverage (measured by first time consultation); protection against pertussis, tetanus and diptheria (measured by third dose of DPT vaccination); and first time antenatal care consultations. Variables were collected for three years (1985-7). (Although not explicitly defined in the study, it appears that the population were grouped into five strata, based on residence/municipality.)

Prior to decentralisation, the pattern of distribution of health service indicators between different strata of the non-insured population in both states tended to be equitable. The most needy received significantly more general and first-time consultations per capita. Oaxaca continued to distribute services on a similar basis, although with a significant decrease in volume across all indicators - due to the economic crisis (table 12). "The negative effects of the economic crisis along the five strata were thus assimilated with equality in the state that remained centralised." By way of contrast, there was a significant shift in the pattern of distribution of services between strata in Guerrero after decentralisation. In spite of the economic crisis, the state as a whole received more services; yet all of the increments benefited the top two strata living in the most important cities and tourist areas of the state. The middle strata experienced little change either way. The two poorest strata, however, suffered a significant drop in services - especially in the interruption of DPT vaccination. These people were mainly Indians and peasants living in small dispersed settlements. The study attributes this regressive shift to the planning and resource allocation policies followed under decentralisation. The tourist resorts of Guerrero received additional resources from the center, by-passing state level negotiations. The middle strata municipalities negotiated on a bilateral basis with state authorities, which seemed to have provided some protection against the economic crisis. The municipalities in the poorest strata signed no decentralisation agreements. Thus the incentives offered by the center to decentralise were better assimilated by the richer and economically more strategic municipalities, in contrast to the net negative effect suffered in the poorest rural areas.

The study concludes that decentralisation in Guerrero increased health differentials between rural and urban areas, and between those sectors engaged in production for local markets and those generating foreign exchange. Inter-district inequities in health care were found to have worsened in Ondo State, Nigeria after decentralisation in 1976 (Iyun,1988). This study compared the distribution of both hospitals and health facilities over time. During the period (1976-85) there were substantial increases in the state's total resources (hospital beds and staff) for health. However some local government areas had an excess share of facilities relative to their population, whilst others had less than their 'fair' share. The favoured districts were described as the 'most viable economically', and included those with large urban centres. The disparities were exacerbated by losses of medical staff from government institutions (because the state could not afford to pay them), and the establishment of increasing numbers of private clinics which tended to be concentrated in the districts already most favoured in terms of public health facilities. Thus the imbalances were regressive in nature, and increased over time. Education Given the serious disparities which affect the educational opportunities of rural, poor and female children in developing countries (HDR 1991 tables 9-10), the enhancement of equity is a foremost concern. There is a vital need for increased resource provision directed towards those districts and for those students who are currently disadvantaged. This argues for strong redistributive measures in educational policy - which can only be fully undertaken by the national government. In favour of decentralisation is the possibility that the demand for education by traditionally excluded groups may be promoted. Yet there is considerable evidence that decentralisation of responsibility for education tends to exacerbate, or at least not reduce, existing inequalities especially between regions. It is a difficult counterfactual question as to whether the expansion of educational opportunities under a decentralised regime, albeit on an uneven basis, is better from the perspective of aggregate inequality, than the situation which would pertain under centralisation. The following paragraphs present the available evidence on these points.

There is strong evidence that the Harambee schools, which proliferated during the 1970s and represented 73% of all secondary schools by 1987, exacerbate interdistrict inequalities in Kenyan education (Mwira 1990). This occurs in several ways. Harambee schools are more important in poorer, less economically advanced parts of the country, yet charge higher fees than government schools. The unequal distribution of school places by province and school type is shown in table 13. There has also been a polarisation of the school system in terms of quality. Using the public examinations as a gauge, government schools and Harambee are on either of the extreme ends of the scale, the former achieving far better results. The consequences of this are reflected in post-secondary enrolment by school type i.e. the fact that Harambee students are far less likely to proceed onto university

(table 14). The factors underlying poor examination performance are inadequate teaching and physical facilities. Few Harambee schools teach the natural and physical sciences, for example, because they cannot afford the equipment. These disparities also have a intra-district gender dimension (see table 12) - because the government built more boys' than girls schools, girls are forced to seek alternatives and are more likely than boys to enrol in Harambee schools. In 1983, some aspects of the recurrent costs of primary education in Zambia were transferred from Lusaka to the district level and to parents. Central education expenditures were basically confined to teachers salaries. Silanda and Tuijnman (1989) found that increased local financing was a major cause of inter-district disparities in total recurrent expenditure per pupil (see table 15). There was a "striking degree of inequality between regions in terms of the funds available for the financing of non-salary components of the recurrent budget. Parental expenditures per pupil on the purchase of learning materials was more than twice as high in three regions (Lusaka, Copperbelt and Central), as in the remaining six - which largely reflected "general rural-urban, socio-economic disparities" (p.13). The authors concluded that the "decentralisation of responsibility for the financing of primary education may run counter to policies emphasising the provision of equal access to a uniformly good primary level education throughout the country" (p.16). Samoff (1990) argues that when local autonomy in education is enhanced, efforts to reduce regional inequalities are undermined. He illustrates this with the case of the 'bush schools' in Kilimanjaro, a better off region in Tanzania. These private secondary schools proliferated in the mid-1970s as government schools were increasingly unable to meet the demand for secondary schooling. They were typically established by a local coalition which included church personnel, government officials, party leaders and other community representatives (often village headmen). Financial assistance comes from the church and local contributions. Once opened, the schools are 'quasi-public' - being subject to Ministerial regulations and funded by a local school tax (collected by village governments). Whilst these schools expanded educational opportunities in Tanzania the initiatives "generally served to thwart national redistributive and equalisation policies" (p.11). The author reports that the representatives of relatively disadvantaged regions preferred greater centralisation, whilst Kilimanjaro leaders seeking to limit redistribution advocated local autonomy. This case may be contrasted to the Kenyan situation (see above), where local school initiatives were of poorer quality than the public system. In India, the distribution of education expenditure between the center and the states is determined by the planning commission (re development expenditure) and the finance commission (for non-plan, statutory and maintenance expenditure). The formulae used by the federal agencies contain a mixture of criteria - 'neutral' (population), progressive (per capita income), and regressive (state of derivation) (Hinchcliffe 1989). However the actual pattern of plan resources does not indicate any strong rationale behind interstate allocation neither state domestic product per capita, nor education development (enrolments etc.) are explanatory factors. Regarding non-plan expenditures from finance

commissions, educationally advanced states receive larger allocations. Whilst this may be because of the larger education systems to be maintained by these states, some have blamed arbitrary discriminatory policies. Tilak (1989) concluded that "on the whole, neither the planning commission nor the finance commission has been able to introduce progressivity into transfers of resources to the states for the education sector. The whole mechanism of federal fiscal transfers has tended to work to the detriment of the weaker states", and that interstate inequalities (measured by the coefficient of variation in the expenditure on education) have correspondingly increased (p.476). Other studies have similarly suggested that federal transfers have not reduced interstate inequalities in India (Hinchcliffe 1989). The reason for this, according to Tilak (1989 p.476), is that "essentially all basic policy decisions in education are political in character. Resource allocation is not exempt". More vocal states, and states having a ruling political party in power identical with, or supportive of, the ruling party at the center are favoured in the process. "In short, the model that best explains allocation of resources by the center to the states for education may be a political model" (id.).

A recent study of decentralisation in education noted that in Eastern Nigeria, wealthier communities were found to respond most strongly to government incentives to build new schools (Winkler, 1989). U.S. evidence suggests that substantial disparities can arise in a decentralised system. The New York State Constitution requires the state legislature to "provide for the maintenance and support of a system of free common schools, wherein all the children of this state may be educated". At the same time however, local taxes constitute about 45% of the total school finances. In Long Island, recurrent expenditure per pupil is $5178 in the poorer districts, compared to $10 529 in the wealthy districts of Great Neck, a wealthier district (NYT, 16/5/91). This is despite the fact that state aid policy is, at least partly, redistributive. Great Neck receives only $360 per pupil, compared to Roosevelt's $2576. Even if the wealthier districts received no state subsidies at all, their capacity to generate revenue still vastly exceeds that of the poorer schools. Thus decentralisation appears to result in large inequities in school funding. Would a centralised system enhance equity? This would partly depend upon the ability of the state to raise the same resources which are currently being collected. There is clearly ability to pay in the wealthier districts. There must also be willingness to contribute to the education of pupils outside the local community. The centralised educational policies of various Latin American countries allowed a massive expansion of education. The combined primary and secondary school enrolment rates have increased from 68% in 1970, to 84% in 1987 (HDR 1991 table 24). Despite this formidable achievement there is said to be a general dissatisfaction with the model pursued, partly due to the expected (but unfulfilled) democratic and equity effects of education policies (Cassusus, 1990, p.12). Wide socio-economic disparities and the exclusion of marginal sectors together with scarcity of materials and teachers are reflected in the highly uneven distribution of

achievements across schools. Statistics show that school failure, expressed in terms of repetition, drop-out and over-age, are mainly concentrated in rural and urban marginal sectors (id.). Hinchcliffe (1989) concluded from a review of financing education in six federal countries (three developing - Brazil, India and Nigeria), that spatial equalisation efforts succeed where the emphasis is on block or general revenue grants, rather than sector or project specific allocations. Redistributive financing is an important, but not the sole, means to enhance educational equity. Other aspects of schooling may significantly impact upon the enrolment and attendance of children from poor families. These may be most appropriately set at the local level.

In Bangladesh, flexible scheduling is one of the key strategies used to increase the school participation of girls and children from poor families. BRAC has established a system of schooling for 8-10 year olds which is open only to the children of landless, illiterate parents. 63% of the pupils are girls - the target is 70%. BRAC is currently planning to establish another system of schools, only for girls, between 13-15 years of age.

The Taxation Effect The system of taxation can have a significant impact on aggregate inequality, through differing tax burdens among income groups in the economy. It has been said that fairness in taxation is especially important where the level of (average) income is low and its distribution is highly skewed (Musgrave 1983). Taxes do take a relatively large share of even very low incomes in most developing countries (Bird and Miller 1990). National and local taxes may well vary in their impact upon intra- and inter-district inequality, given the different bases and tax rates which are utilised. Unfortunately there is little evidence as to whether local systems of taxation are relatively more or less progressive than those at the national level. This is not very suprising, given the considerable difficulties involved in estimating the incidence of the tax burden in developing countries. Basic data is required as to the distribution of income by types of earning, expenditure patterns by income levels and sectors, and the distribution of property and assets. Any analysis will be based on assumptions about the incidence of specific taxes. Nor can equity be evaluated without reference to relative tax compliance especially where the rich enjoy greater opportunities to evade payment. Despite the empirical problems, it has been said that national taxation policies in developing countries are generally found to be "not very regressive" (Bird and Miller 1990). Typically, the main sources of national revenue are levied on (in descending order): foreign trade, income (corporate and personal), sales and resource extraction. As in industrial countries, the weight of opinion as to the incidence of corporations tax in developing countries is inconclusive (Lent 1990). The basic unknown is whether the tax

is shifted in the long run, either back to labour (in the form of lower real wages), or forward to consumers. The incidence of indirect sales taxes - on tobacco, alcohol and oil - appears to vary sharply from country to country, depending on the specifics of local consumption and tax structures (Bird and Miller 1990). Whilst some have concluded that "in most cases the existing structure of indirect taxation is markedly regressive" (Chenery 1974), others argue for detailed quantitative study rather than "sweeping generalisation" (Bird and Miller 1990). Clearly distinctions must be drawn amongst different socio-economic groups - poor rural, largely subsistence households are likely to have different expenditure patterns than the urban poor. The nature of local taxes tends to suggest that the burden is not very regressive. The main progressive revenue sources available to local governments in developing countries are automobile use and ownership, and real property. Taxes on motoring - vehicles and gasoline - may be progressive insofar as the demand tends to be highly income elastic (Cnossen 1990). Bahl and Linn (forthcoming) found that of 19 studies in thirteen developing countries, most found the property tax to be generally or very progressive although the actual distributive impact of such taxes obviously depends largely upon its details (Bird and Miller 1989) - for example the treatment of rental versus owneroccupied, of land versus improvements, and so on. On the negative side, the multitude of local revenue sources, levied on entertainment, sumptuary items, small businesses and so on, may be regressive in net impact.

In Indonesia, the equity effects of existing local taxes have been seen as follows (Devas 1988). Motor vehicle taxes are relatively equitable, since car ownership is limited to upper income groups. Taxes on entertainment are reasonably equitable, because the tariffs vary according to type of entertainment, with traditional performances having the lowest rates. Taxes on restaurants and hotels are progressive, since the rich are likely to spend proportionately more on these items, and small establishments and itinerant vendors are exempt. The effects of the registration taxes levied on businesses are uncertain, depending on who the final consumers are and what proportion of the tax is passed on. Finally, the tax on non-motorised vehicles can be regarded as regressive, since users of bicycles and becaks are likely to be among the poorer members of the community.

7. Obstacles The actual process of implementation and the obstacles confronted thereunder raises issues which do not always directly impact upon human development, yet nonetheless underlie the whole discussion of decentralisation. This survey would be incomplete without an appraisal of the operation and significance of these obstacles. Indeed this is regarded by many commentators as the most important issue to address (e.g. Rondinelli, 1984). It will have emerged from the foregoing that decentralised systems of government should be appraised in terms of actual operation, rather than ostensible aims and designs.

This is especially relevant in deeming particular decentralisation efforts as representing either deconcentration or devolution. Significant practical problems can inhibit the implementation of government efforts to decentralise. They range from the political - central decision makers are likely to resist any diminution of their powers - to the practical, especially shortages of resources and the difficulties arising due to weak local administrative capacity. Ministers of finance in the national government may oppose the transfer of revenue sources to decentralised units, whilst reliance on central funding can mean chronic delays, revenue shortfalls and reduced local autonomy. For example, it has been reported that the bureaucratic obstacles to decentralisation in Nepal were so serious that local officials were frequently required to travel to Kathmandu in order to expedite the release of central grants (Bienen et al 1990). Political constraints "Where decentralisation is proposed seriously, it is likely to be a focus for political confrontation, precisely because it has to do with specifying who may participate legitimately in decision making and whose interests will have highest priority" (Samoff 1990, p.524). It is predictable that change will be resisted by those who perceive their vested interests threatened by a new structure of governmental arrangements. This, coupled with bureaucratic inertia, can constitute formidable obstacles to the implementation of decentralisation reforms. Political will at the highest level may not, per se, be sufficient to overcome such obstacles. A number of authors criticise the assumption that national governments are "unitary, neutral actors". McGinn and Street (1986) argue that policy makers evaluate organisational reforms in terms of the impact upon their own power and control over the use of public resources - so that different actors react differently to the same policy.

Since 1983 the Rawlings government in Ghana has been pursuing a decentralisation proposal that would transfer virtually all ministerial activities through the regions to the localities. In 1988, District Assemblies were elected which were supposed to take command of district-level administrations onto which the resources and expertise of central ministries were to be devolved. Olowu (1990) reported that progress has been very slow because of powerful forces within the administration against the plan; the distrust of elected councillors by military and civil servants alike; and the widespread feeling that decentralisation was a diversion from national economic policies. In the meantime, the old structures have continued through selected management committees presiding over deconcentrated administration at regional, district and local levels. An IRD project in Peru in the early 1980s which sought to combine regional spatial planning with decentralised administration was "ultimately undermined by

political obstacles and central government controls" (Rondinelli and Wilson 1987 p. 348). The most serious constraints stemmed from continuing central controls on local planning and resource allocation. This "not only inhibited the local development corporations from responding to local needs and conditions, but subjected them to manipulation by the dominant national political party" (ibid p.353). The highly centralised control meant a vacuum of information about local conditions which "in turn generated apathy toward regional problems and this, combined with the lack of power or resources at the local level, created pervasive inertia against acting promptly on regional plans or local development proposals. In rare cases expeditious action occurred when local political groups bypassed departmental channels and went directly to the central ministries or influential congressmen in Lima. The study concluded that the potential for decentralisation depended crucially upon increasing political commitment to decentralised planning and decision making (ibid. p. 356). Bureaucratic obstacles Beyond the political tensions involved in decentralisation attempts, there often arise significant problems related to the nature of the bureaucracy - its structure, technical competence and resource base. The creation of new systems of local government requires considerable administrative skill and resources. There may be basic problems with respect to personnel needs, attitudes (held both locally and centrally), and training in local government. In a number of cases, the central government has retained effective control over local personnel through recruitment, promotion, payment and so on. Although this may help to alleviate skill and resource shortages at the local level, it leads to accountability and loyalties which lead upward to the capital, rather than to local constituents. This may undermine local autonomy, and create an obstacle to the implementation of decentralising reforms.

In Indonesia, it has been observed that the structure and regulations governing the civil service severely handicap efforts to develop a local civil service adequate to meet the demands of decentralisation (King, 1988). Substantial deconcentration of development activity through INPRES programmes was found to have been undermined by centralising civil service reforms (as well as financing arrangements, see ** above). In 1985 over three-quarters of employees working in regional governments were seconded central employees. Local staff available for decentralised management was subject to cumbersome and centralised control. Further, the structure of incentives encouraged a "brain drain" from the regions to Jakarta. In Bangladesh, although the services of civil servants has been formally transferred to the upazilas, their appointment, training, transfer, promotion and discipline all remain subject to national government control (Khan, 1987, op cit). In Botswana the recruitment, posting, training, promotion and discipline of local authority staff is generally the responsibility of the centrally directed Unified

Local Government Service (ULGS). The ULGS staff at headquarters has been strongly criticised by the District Councils for not being sufficiently aware of, and responsive to, local problems and needs (Tordoff 1988 p.198). Staff shortages at the local level appear to be a crucial constraint in many developing countries. An IMF survey conducted in 1982 showed that 57% of all government employment in industrial countries is at the 'local' (as opposed to the central) level, compared to only 15% for developing countries - Africa averaging only 6%, Latin America 21% and Asia 37% (World Bank 1983). This can be seen to reflect personnel shortages as well as the fact that local governments account for a greater share of public expenditure in industrial, relative to developing, countries.

In Nigeria, there is reportedly a wide gap between the staff required by local governments and the actual number employed by them. Local governments have not been able to attract and retain qualified staff and attempts by state and federal governments to train local government staff have not overcome this problem (Nwankwo 1984). Rondinelli et al (1983) emphasise the "crucial effects of shortages of trained staff on the success of decentralisation" (p.69). There are many examples of local governments with limited capacity. In the Chilean context of a relatively thoroughgoing decentralisation, the limited capacity of local officials has confined their work to mainly administering central directives: local initiatives have been rare (Racynski 1987). In Kenya the vast majority of skilled technicians and managers have been concentrated in Nairobi. The Sudan has suffered severe skill shortages, even within central ministries, which impeded decentralisation and limited the ability of provincial administrators to provide even basic services. In Pakistan, the technical officers available to the markaz councils were quite limited - in that the project managers are often either agricultural technicians or generalist administrators with little experience of district planning and development (Khan 1982).

The Moroccan government recognised the importance of personnel at the local level in creating a new special corps of posts in the local civil service, consisting of centrally trained and paid chief civil servants called 'secretary-generals'. Training centres for local civil servants were greatly expanded (Nellis 1983). In Zambia, administrative capacity to plan and implement substantial development programmes was lacking at the "provincial, let alone district level (and) even in Lusaka". The shortage of qualified staff, the absence of suitable training programmes for district development officers, the unattractiveness of local government employment all contributed to this problem (Rakodi 1988). In Kenya there are several management problems at the local level. Many officers tend to be underqualified for their jobs; in some cases appointment is on the basis of political criteria rather than merit. There are frequently severe gaps in staffing. Some councils have six or seven vacancies at senior levels (clerk, treasurer, surveyor etc.) - so that one officer might be expected to fulfil all of these tasks

over an extended period. In this light it is noted that in the 1984-88 National Plan period, even the more successful councils only managed to carry out about half the projects they had proposed (Wallis 1990). In Botswana the separation of the 'Tribal Administration' from the Village Development Committees is said to have stifled the effectiveness of the latter. Although the traditional institution has been stripped of its statutory powers it remains influential, as "the bulk of the rural people still look to the Chief and ... the traditional hierarchy for advice and guidance on a wide range of issues" (Tordoff 1988). Another example of institutional weaknesses undermining the efficacy of local government comes from Liberia. In the period 1974-1980, a World Bank-funded urban development project in the Liberian capital concentrated on infrastructural improvements. The Monrovia City Council had very limited spending and administrative authority. Local revenue covered only 7% of local expenditure, accounting systems were very weak, and mismanagement occurred frequently. The responsibilities of the MCC were poorly defined and qualified staff were lacking.

The World Bank worked outside of this structure through its "Project Team"; by the end of 1985 the planned civil works were largely completed, which benefited an estimated 130 000 people. However little had been accomplished in the way of institutional reform. The MCC became increasingly dependent upon central funds, as the government was becoming less reliable. One evaluation (by Price Waterhouse) identified the following weaknesses: (i) shortages of professional skills; (ii) responsiveness of the appointed mayor and council to their superiors, rather than local residents; (iii) inadequate revenue and weak fiscal discipline; (iv) uncoordinated development programmes; and (v) failure to enforce local regulations. The role of NGOs and traditional structures It can be argued that human development is such a complex issue that a wide variety of institutions - market, central government and field administration, local government and non governmental organisations - are required. Hyden, for example, writes in the African context that genuine local governments must build upon the wide variety of traditional and modern local associations, because otherwise people will perceive local governments to be an extension of the State. Similarly, Uphoff argues that there should be "multiple channels of action". Since any one channel can become monopolised or blocked, it is said to be better that there be a degree of competition as well as complementarity between various institutions - whether bureaucratic, local government, co-operative, private enterprise, NGO, religious or other. More recently Wunsch and Olowu (1990) have argued along the same lines. The attention accorded to NGOs follows their apparent comparative advantage in areas of priority for human development, given their flexibility and small size, as well as highly motivated personnel (Paul, in Lewis 1988). However rather than attempt to assess the relative performance of NGO vis- -vis that of the various levels of government, the ambit

of discussion is more limited. It investigates the proposition that local, rather than national, governments are better able to create an environment within which the work of NGOs is facilitated. Turning then to evidence of collaborative relationship between government and grassroots organisations, the following points emerge. There is already considerable operational collaboration in both planning and delivery of public goods for human development - e.g. Amul Dairy in India and the Grameen Bank (Paul, op cit). In Indonesia, the national government provides the funds, supplies and technical expertise needed for family planning programmes, whilst community volunteers and village organisations mobilise demand and assist in the distribution of contraceptive supplies. An important related issue is the role of traditional institutions. This has been explored in some detail in the African context. Traditional institutions have generally not been invested with measures of self-governance. In East Africa - Ethiopia, Kenya and Tanzania - and elsewhere governments have not treated them as critical agents in ensuring the appropriateness and sustainability of development initiatives (Harbeson 1990). Sometimes this has been due to their association with the former colonial regime, or with conservative obstacles to reform in the country. A number of writers have emphasised that national governments have neglected and even deliberately undermined traditional self-governing bodies; and that this, in turn, has inhibited the effectiveness of local government. "Rather than build and improve on these institutions, they replaced them with expanded state structures, often with disastrous results. Furthermore, historical and anthropological studies indicate that local self-governing structures are more compatible with several traditional African political systems than the nominally `decentralised' structures created by African states. These traditional self-governing structures (ethnic, town and community unions) have continued to co-exist with the newly created structures in several African countries with very little interaction between them" (Olowu 1990). In Africa traditional institutions were the mechanisms by which the people regulated their social and economic affairs, and exercised and controlled political power. Traditional authority encompassed chiefs and councils of elders and the complex bodies of law which guided their actions; for example, the village councils promulgated and enforced access rules that regulated the balance between livestock, water and forage, and minimised overgrazing, and complex rules of access and use assured (young men across) most of Africa access to land. Since independence, the tendency has been to abolish any legal authority which remained in traditional institutions. Traditional leaders have, as in the colonial era, frequently been 'co-opted', through recruitment into the civil service or pensions. In some cases resistance to such policies has led to bloody violence - as in Southern Sudan, Uganda and Ghana. In

the result, the absence of effective local institutions is said to inter alia threaten common resources and community endeavours. Deforestation and decertification processes in the Sahel have been traced directly to the absence of local institutions able to regulate tree use and livestock foraging. In Botswana, the abolition of local traditional authorities made it far more difficult to solve such problems (Wunsch 1990). Also, as noted above, traditional leaders appear to have been far more effective than government officials in collecting local taxation dues.

Occasional Paper 13 - DECENTRALISATION: A SURVEY OF LITERATURE FROM A HUMAN DEVELOPMENT PERSPECTIVE

6. Equity

The Extent and Impact of Central Grants The Expenditure Effect The Taxation Effect

7. Obstacles

Political constraints Bureaucratic obstacles The role of NGOs and traditional structures

6. Equity Measured in terms of various indicia of human development, there are marked disparities between rural and urban areas, regions, and income groups in developing countries. This has been an important theme of the HDRs. Many examples can be cited (see HDR 1990 pp 29-33; 1991 pp.26-7). Thus the distribution of central grants, public expenditures, and the burden of taxation under a decentralised system are all critical aspects of the present survey. In this section the theoretical arguments and any available empirical evidence are brought to bear upon these issues. It is important to recognise the multi-dimensional nature of equity. In the present context, one must distinguish between inter - district (between different parts of the country), and intra - district (the position of different socio-economic and gender groups within the locality vis- -vis each other) aspects of equity. Improvements or deterioration in these various elements of equity need not coincide. Indeed it appears that while decentralisation can promote equity among different groups within a region, through increased local public expenditure and the wider provision of public goods and services, there is a risk that disparities between districts may worsen due to, for example, different tax bases and inadequate fiscal equalisation. The former tendency points in the direction

of greater decentralisation, whilst the latter suggests that the central government should at least retain a strong redistributive role. There are several critical factors which determine how de/centralisation influences equity. These include central grants, expenditure and taxation effects. It is clear that no uniform generalisations may be forthcoming about the impact of decentralisation on aggregate inequality, since the impact of the various factors may be working in different directions. On balance, the evidence uncovered by the present survey suggests that aggregate inequality is likely to be predominated by worsening inter-regional equity, largely because of the absence of effective central redistributive policies. Intra-district If decentralisation leads to increased local public expenditure and expands the provision of public goods and services, a narrowing of intra-district disparities in access among different socio-economic groups would be expected to follow. Moreover, if decentralisation of decision making is accompanied by increased participation of representatives from a wider variety of political, ethnic and social groups, greater equity in the allocation of public resources is presumed to be more likely (Uphoff and Esman 1974). Whether the latter is likely to eventuate was discussed in section 3 above. The reader will recall that many commentators are pessimistic as to the likelihood of active broad based participation in decision making under decentralisation. According to Hughes (1985), "too frequently the local beneficiaries are not the poor, as much as elite groups and political factions controlling the local authorities" (p.59). On the other hand the empirical evidence brought together on the topic of comparative priorities in the present survey suggests that local governments tend to spend a higher proportion if their budgets on social services, and on more basic levels of provision therein, than do central governments (see Part IV). From a theoretical perspective, inequalities within specific localities could worsen under decentralisation, because the redistributive functions of local government are thereby limited for two related reasons (Oates, 1985): (a) On the taxation side, "sorting out along Tiebout lines" will imply relatively little scope for redistribution from the wealthy to the poor within jurisdictions; and (b) Local redistributive expenditures will be constrained by the fear of attracting the mobile poor with relatively generous local support programmes. This suggests that the actual scope for assistance to the poor within the locality will be more circumscribed under a relatively decentralised fiscal system. Inter - district The impact of decentralisation upon spatial disparities in human development is potentially two-fold. Decentralisation, even in terms of its more limited mode

(deconcentration), can reduce spatial inequalities in access to basic education and health care, clean water, decent income earning opportunities and so on, through the wider provision of public goods and services. On the other hand however, if decentralisation extends to the mobilisation of public resources (i.e. revenues as well as expenditures), existing disparities are likely to be exacerbated. This follows from the observation that the poorer regions of the country will have correspondingly lower fiscal potentials. In practice, there is evidence which reflects the operation of both scenarios. A number of studies support the proposition that inter-district disparities may be reduced through the wider provision of public goods and services. Rondinelli et al (1983) list among the positive results of decentralisation that "the access of people living in previously neglected rural regions and local communities to central government resources and institutions has increased, if only incrementally, in most developing countries that have decentralised" (p.50).

In Morocco, the Local Authorities Development Fund (created as part of the decentralisation effort) distributed funds throughout the country on a far more equitable basis than its predecessor, the Communal Supply Fund, which had mainly been drawn upon by richer, urban communes (Cheema 1983). Maro (1990) found that in Tanzania, data on clean water, health facilities, primary education and roads showed that decentralisation had led to: (i) increased government expenditure on these services, which resulted in increased provision throughout the rural areas, and (ii) a drastic reduction in user-distance to basic social services, tending toward uniform access to services in all districts, and therefore a reduction in inter-district disparities.

The Extent and Impact of Central Grants The center has a potentially powerful redistributive role, in collecting revenue on the basis of ability to pay, and allocating funds on the basis of need. The latter can be done either through a bias towards poorer areas (tackling inter-district inequality), or via specific grants which tend to benefit the less well off (e.g. for primary education), thereby enhancing intra-district equity. The available evidence relates mainly to the impact of grants upon inter-district equality. It has been widely argued by writers on fiscal federalism (e.g. Musgrave 1959; Oates 1974) that the national government has the primary responsibility for initiating and coordinating policies so as to achieve horizontal equity - i.e. that policies for redistribution are most appropriately set and implemented centrally. This is partly because there should be a single level of government which has the role of determining the extent of redistribution, lest different decisions made by various levels of government undo the redistributive actions of the other. Moreover, the marked spatial variations in resource availability point to the need for the center to adopt the predominant role in redistribution

(as in, for example, Indonesia - see case study). There is also the 'Tiebout effect' of population mobility, which limits the feasible extent of variance between independent redistributive measures at the local level (Helm and Smith 1987), and suggests the need for a strong central role in promoting an equitable distribution of public resources. (It is noted that the need for redistributive grants conflicts, to some extent, with the presumed need for a substantial degree of financial autonomy, discussed in section 2 above.) The central government may undertake substantive schemes of redistribution through grants which favour local governments in poorer parts of the country. In practice however, it appears that the prevailing fiscal arrangements under decentralisation - by design or otherwise - generally fail to enhance inter-district equity.

In Chile, the system of central grants was altered from one which redistributed in favour of poorer regions to one which made allocations on a per capita basis. Inadequate central financing failed to cover the cost of service provision (as it was supposed to), so that poorer municipalities could only afford inferior services thereby worsening existing inequalities between regions. Zimbabwe proceeded on similar lines to Chile, in that many recurrent central grants are allocated on a per capita basis with no consideration of community fiscal capacity (Winkler 1989). In India, it has been said that federal education grants are made primarily on political grounds and serve only to exacerbate spending disparities (Tilak, 1989). In Bangladesh, the fiscal system transfers resources from the urban to the rural sector (where 85% of the population lives). At the same time however, development fund grants are allocated on a per capita basis without regard to need, and the infrastructure grants appear to be distributed on an ad hoc basis (Bahl, in Schroeder, 1989). In Botswana, where the District Councils depend upon central transfers for over 80% of total recurrent expenditure, the system of resource allocation is described as "haphazard", resulting in the "uneven provision of local government services in different parts of Botswana" (Tordoff 1988 p.197). By way of contrast there is some evidence which reflects attempts by national governments to alleviate the inter-district equity problem, through grants which offset differentials in local tax bases.

In Uttar Pradesh, grants are distributed to the districts on a criterion of "backwardness" - taking total population, number of scheduled castes (who tend to be landless) and development of infrastructure (roads and electrification) in the district. Under this arrangement, which was introduced with decentralisation, the backward districts get more than in the past, and inter-district disparities have tended to be reduced (Sanwal, 1987). In Nigeria, constitutional changes over time have led to an equalising trend in federal transfers. Until 1978 transfers were made on the basis of derivation, as well as other criteria which focused more upon redistribution. In successive constitutional amendments, the latter gained in importance. The heavily populated

and poorer northern states have been able to reduce the influence of the derivation principle in transfers and extend, first, the population criterion and, more recently, the "additional need" factor. Thus, new formulas have further increased the importance of the fiscal equalisation principle (Hinchcliffe 1989 p.445). In Brazil the 1967 tax reform led to the redistribution of public funds from richer to poorer regions (Graham 1987). The basic formula used by the federal government weighted the amount received according to the level of development. For every 100 cruzeiros collected by the federal government in taxes set aside by law for transfer to the states, the North received back 210 cruzeiros in 1976; the states of the South and Southeast regions receive less than 25 cruzeiros for each 100 they collect. In the education sector, the World Bank found that the formula for distributing federal education moneys to the states in Brazil is highly redistributive. Similarly, Eastern Nigeria and Kenya have taken actions to redistribute government grants in favour of poorer community schools (Winkler, 1989).

The Expenditure Effect Decentralisation may enhance equity through increased public expenditure on those areas of most benefit to the poor. Improved provision and utilisation of basic social and economic services can follow. This relates to the discussion of relative priorities in Part IV below. In the following sections, the expenditure effect of decentralisation is appraised from the perspective of equity, both intra and inter-district, in the specific contexts of health and education. Different characteristics and concerns tend to be relevant to the each sector. In practice it is found that decentralisation does tend to expand the provision of basic social and economic infrastructure. Yet inequalities persist, and are sometimes exacerbated. This may be due to disparities in resource allocation or in utilisation patterns, or in the ability to respond to incentives under a decentralised system. It is a moot point as to whether this leads to a more or less equitable situation overall than that which might have existed under a centralised system. Health care Equity is an important consideration in the health context, where access to services (and indicators) varies markedly across different regions, income groups and sexes (HDR table 9). By increasing access throughout the country, decentralisation may help to reduce these disparities. However the limited evidence available on this point tends to suggest that the expansion of health resources under a decentralised regime does not tend to be equitably distributed, even within regions.

The most positive example of an equitable expansion of services across regions under a deconcentrated system is provided by Tanzania. Maro (1990) found that there was a clear bias in favour of regions that had fewer facilities prior to decentralisation, such as Dodoma, Coast and Mwanza. The result was a dramatic

increase in the provision of health facilities. In 1972, only one in eight people lived within five kilometres of a hospital; in 1980, 60% of the villages either had a health facility or were within five kilometres of one. In Indonesia, between 1969 and 1983, in tandem with increasing decentralisation of public expenditures, there was a rapid expansion of health services. More recently the Posyandu (community-based integrated health post, mostly concerned with child survival) has been extended to all parts of the country. Currently, the Kecamatan (sub-district) health center provides the administrative framework within which most basic health care activities take place. Yet despite this substantial deconcentration of health operations, there exist large disparities in utilisation patterns among different groups in Indonesian society which impact adversely upon intra-district equity. It has been found that, on average, the insured population (civil servants and their families) used the health services four times more frequently than the rest of the population. In one provincial study, the top income decile made up one-third of all hospital inpatients, one-half of hospital outpatients, and one-quarter of all health center visitors (Gish et al 1988).

A recent critique of the decentralisation of health services in Mexico provides an interesting counterfactual-type study (Gonzalez-Block, 1989). It also introduced the important variable of macro-economic conditions. The study covered two states in the poorer parts of the country, one of which decentralised to a significant extent (Guerrero), whilst the other did not (Oaxaca). It found that intermunicipal disparities worsened in decentralised Guerro. The statistics used to indicate health need satisfaction were: the number of general consultations; extension of service coverage (measured by first time consultation); protection against pertussis, tetanus and diptheria (measured by third dose of DPT vaccination); and first time antenatal care consultations. Variables were collected for three years (1985-7). (Although not explicitly defined in the study, it appears that the population were grouped into five strata, based on residence/municipality.)

Prior to decentralisation, the pattern of distribution of health service indicators between different strata of the non-insured population in both states tended to be equitable. The most needy received significantly more general and first-time consultations per capita. Oaxaca continued to distribute services on a similar basis, although with a significant decrease in volume across all indicators - due to the economic crisis (table 12). "The negative effects of the economic crisis along the five strata were thus assimilated with equality in the state that remained centralised." By way of contrast, there was a significant shift in the pattern of distribution of services between strata in Guerrero after decentralisation. In spite of the economic crisis, the state as a whole received more services; yet all of the increments benefited the top two strata living in the most important cities and tourist areas of the state. The middle strata experienced little change either way. The two poorest strata, however, suffered a significant drop in services - especially in the

interruption of DPT vaccination. These people were mainly Indians and peasants living in small dispersed settlements. The study attributes this regressive shift to the planning and resource allocation policies followed under decentralisation. The tourist resorts of Guerrero received additional resources from the center, by-passing state level negotiations. The middle strata municipalities negotiated on a bilateral basis with state authorities, which seemed to have provided some protection against the economic crisis. The municipalities in the poorest strata signed no decentralisation agreements. Thus the incentives offered by the center to decentralise were better assimilated by the richer and economically more strategic municipalities, in contrast to the net negative effect suffered in the poorest rural areas. The study concludes that decentralisation in Guerrero increased health differentials between rural and urban areas, and between those sectors engaged in production for local markets and those generating foreign exchange. Inter-district inequities in health care were found to have worsened in Ondo State, Nigeria after decentralisation in 1976 (Iyun,1988). This study compared the distribution of both hospitals and health facilities over time. During the period (1976-85) there were substantial increases in the state's total resources (hospital beds and staff) for health. However some local government areas had an excess share of facilities relative to their population, whilst others had less than their 'fair' share. The favoured districts were described as the 'most viable economically', and included those with large urban centres. The disparities were exacerbated by losses of medical staff from government institutions (because the state could not afford to pay them), and the establishment of increasing numbers of private clinics which tended to be concentrated in the districts already most favoured in terms of public health facilities. Thus the imbalances were regressive in nature, and increased over time. Education Given the serious disparities which affect the educational opportunities of rural, poor and female children in developing countries (HDR 1991 tables 9-10), the enhancement of equity is a foremost concern. There is a vital need for increased resource provision directed towards those districts and for those students who are currently disadvantaged. This argues for strong redistributive measures in educational policy - which can only be fully undertaken by the national government. In favour of decentralisation is the possibility that the demand for education by traditionally excluded groups may be promoted. Yet there is considerable evidence that decentralisation of responsibility for education tends to exacerbate, or at least not reduce, existing inequalities especially between regions. It is a difficult counterfactual question as to whether the expansion of educational opportunities under a decentralised regime, albeit on an uneven basis, is better from the perspective of aggregate inequality, than the situation which would pertain under centralisation. The following paragraphs present the available evidence on these points.

There is strong evidence that the Harambee schools, which proliferated during the 1970s and represented 73% of all secondary schools by 1987, exacerbate interdistrict inequalities in Kenyan education (Mwira 1990). This occurs in several ways. Harambee schools are more important in poorer, less economically advanced parts of the country, yet charge higher fees than government schools. The unequal distribution of school places by province and school type is shown in table 13. There has also been a polarisation of the school system in terms of quality. Using the public examinations as a gauge, government schools and Harambee are on either of the extreme ends of the scale, the former achieving far better results. The consequences of this are reflected in post-secondary enrolment by school type i.e. the fact that Harambee students are far less likely to proceed onto university (table 14). The factors underlying poor examination performance are inadequate teaching and physical facilities. Few Harambee schools teach the natural and physical sciences, for example, because they cannot afford the equipment. These disparities also have a intra-district gender dimension (see table 12) - because the government built more boys' than girls schools, girls are forced to seek alternatives and are more likely than boys to enrol in Harambee schools. In 1983, some aspects of the recurrent costs of primary education in Zambia were transferred from Lusaka to the district level and to parents. Central education expenditures were basically confined to teachers salaries. Silanda and Tuijnman (1989) found that increased local financing was a major cause of inter-district disparities in total recurrent expenditure per pupil (see table 15). There was a "striking degree of inequality between regions in terms of the funds available for the financing of non-salary components of the recurrent budget. Parental expenditures per pupil on the purchase of learning materials was more than twice as high in three regions (Lusaka, Copperbelt and Central), as in the remaining six - which largely reflected "general rural-urban, socio-economic disparities" (p.13). The authors concluded that the "decentralisation of responsibility for the financing of primary education may run counter to policies emphasising the provision of equal access to a uniformly good primary level education throughout the country" (p.16). Samoff (1990) argues that when local autonomy in education is enhanced, efforts to reduce regional inequalities are undermined. He illustrates this with the case of the 'bush schools' in Kilimanjaro, a better off region in Tanzania. These private secondary schools proliferated in the mid-1970s as government schools were increasingly unable to meet the demand for secondary schooling. They were typically established by a local coalition which included church personnel, government officials, party leaders and other community representatives (often village headmen). Financial assistance comes from the church and local contributions. Once opened, the schools are 'quasi-public' - being subject to Ministerial regulations and funded by a local school tax (collected by village governments). Whilst these schools expanded educational opportunities in Tanzania the initiatives "generally served to thwart national redistributive and equalisation policies" (p.11). The author reports that the representatives of relatively disadvantaged regions preferred greater centralisation, whilst

Kilimanjaro leaders seeking to limit redistribution advocated local autonomy. This case may be contrasted to the Kenyan situation (see above), where local school initiatives were of poorer quality than the public system. In India, the distribution of education expenditure between the center and the states is determined by the planning commission (re development expenditure) and the finance commission (for non-plan, statutory and maintenance expenditure). The formulae used by the federal agencies contain a mixture of criteria - 'neutral' (population), progressive (per capita income), and regressive (state of derivation) (Hinchcliffe 1989). However the actual pattern of plan resources does not indicate any strong rationale behind interstate allocation neither state domestic product per capita, nor education development (enrolments etc.) are explanatory factors. Regarding non-plan expenditures from finance commissions, educationally advanced states receive larger allocations. Whilst this may be because of the larger education systems to be maintained by these states, some have blamed arbitrary discriminatory policies. Tilak (1989) concluded that "on the whole, neither the planning commission nor the finance commission has been able to introduce progressivity into transfers of resources to the states for the education sector. The whole mechanism of federal fiscal transfers has tended to work to the detriment of the weaker states", and that interstate inequalities (measured by the coefficient of variation in the expenditure on education) have correspondingly increased (p.476). Other studies have similarly suggested that federal transfers have not reduced interstate inequalities in India (Hinchcliffe 1989).

The reason for this, according to Tilak (1989 p.476), is that "essentially all basic policy decisions in education are political in character. Resource allocation is not exempt". More vocal states, and states having a ruling political party in power identical with, or supportive of, the ruling party at the center are favoured in the process. "In short, the model that best explains allocation of resources by the center to the states for education may be a political model" (id.).

A recent study of decentralisation in education noted that in Eastern Nigeria, wealthier communities were found to respond most strongly to government incentives to build new schools (Winkler, 1989). U.S. evidence suggests that substantial disparities can arise in a decentralised system. The New York State Constitution requires the state legislature to "provide for the maintenance and support of a system of free common schools, wherein all the children of this state may be educated". At the same time however, local taxes constitute about 45% of the total school finances. In Long Island, recurrent expenditure per pupil is $5178 in the poorer districts, compared to $10 529 in the wealthy districts of Great Neck, a wealthier district (NYT, 16/5/91). This is despite the fact that state aid policy is, at least partly, redistributive. Great Neck receives only $360 per pupil, compared to Roosevelt's $2576. Even if the wealthier districts received no state subsidies at all, their capacity to generate revenue still vastly exceeds that of the poorer schools. Thus decentralisation appears to result in large inequities in school funding. Would a centralised system

enhance equity? This would partly depend upon the ability of the state to raise the same resources which are currently being collected. There is clearly ability to pay in the wealthier districts. There must also be willingness to contribute to the education of pupils outside the local community. The centralised educational policies of various Latin American countries allowed a massive expansion of education. The combined primary and secondary school enrolment rates have increased from 68% in 1970, to 84% in 1987 (HDR 1991 table 24). Despite this formidable achievement there is said to be a general dissatisfaction with the model pursued, partly due to the expected (but unfulfilled) democratic and equity effects of education policies (Cassusus, 1990, p.12). Wide socio-economic disparities and the exclusion of marginal sectors together with scarcity of materials and teachers are reflected in the highly uneven distribution of achievements across schools. Statistics show that school failure, expressed in terms of repetition, drop-out and over-age, are mainly concentrated in rural and urban marginal sectors (id.).

Hinchcliffe (1989) concluded from a review of financing education in six federal countries (three developing - Brazil, India and Nigeria), that spatial equalisation efforts succeed where the emphasis is on block or general revenue grants, rather than sector or project specific allocations. Redistributive financing is an important, but not the sole, means to enhance educational equity. Other aspects of schooling may significantly impact upon the enrolment and attendance of children from poor families. These may be most appropriately set at the local level.

In Bangladesh, flexible scheduling is one of the key strategies used to increase the school participation of girls and children from poor families. BRAC has established a system of schooling for 8-10 year olds which is open only to the children of landless, illiterate parents. 63% of the pupils are girls - the target is 70%. BRAC is currently planning to establish another system of schools, only for girls, between 13-15 years of age.

The Taxation Effect The system of taxation can have a significant impact on aggregate inequality, through differing tax burdens among income groups in the economy. It has been said that fairness in taxation is especially important where the level of (average) income is low and its distribution is highly skewed (Musgrave 1983). Taxes do take a relatively large share of even very low incomes in most developing countries (Bird and Miller 1990). National and local taxes may well vary in their impact upon intra- and inter-district inequality, given the different bases and tax rates which are utilised. Unfortunately there is little evidence as to whether local systems of taxation are relatively more or less progressive than those at the national level. This is not very suprising, given the considerable difficulties involved in estimating the incidence of the tax burden in

developing countries. Basic data is required as to the distribution of income by types of earning, expenditure patterns by income levels and sectors, and the distribution of property and assets. Any analysis will be based on assumptions about the incidence of specific taxes. Nor can equity be evaluated without reference to relative tax compliance especially where the rich enjoy greater opportunities to evade payment. Despite the empirical problems, it has been said that national taxation policies in developing countries are generally found to be "not very regressive" (Bird and Miller 1990). Typically, the main sources of national revenue are levied on (in descending order): foreign trade, income (corporate and personal), sales and resource extraction. As in industrial countries, the weight of opinion as to the incidence of corporations tax in developing countries is inconclusive (Lent 1990). The basic unknown is whether the tax is shifted in the long run, either back to labour (in the form of lower real wages), or forward to consumers. The incidence of indirect sales taxes - on tobacco, alcohol and oil - appears to vary sharply from country to country, depending on the specifics of local consumption and tax structures (Bird and Miller 1990). Whilst some have concluded that "in most cases the existing structure of indirect taxation is markedly regressive" (Chenery 1974), others argue for detailed quantitative study rather than "sweeping generalisation" (Bird and Miller 1990). Clearly distinctions must be drawn amongst different socio-economic groups - poor rural, largely subsistence households are likely to have different expenditure patterns than the urban poor. The nature of local taxes tends to suggest that the burden is not very regressive. The main progressive revenue sources available to local governments in developing countries are automobile use and ownership, and real property. Taxes on motoring - vehicles and gasoline - may be progressive insofar as the demand tends to be highly income elastic (Cnossen 1990). Bahl and Linn (forthcoming) found that of 19 studies in thirteen developing countries, most found the property tax to be generally or very progressive although the actual distributive impact of such taxes obviously depends largely upon its details (Bird and Miller 1989) - for example the treatment of rental versus owneroccupied, of land versus improvements, and so on. On the negative side, the multitude of local revenue sources, levied on entertainment, sumptuary items, small businesses and so on, may be regressive in net impact.

In Indonesia, the equity effects of existing local taxes have been seen as follows (Devas 1988). Motor vehicle taxes are relatively equitable, since car ownership is limited to upper income groups. Taxes on entertainment are reasonably equitable, because the tariffs vary according to type of entertainment, with traditional performances having the lowest rates. Taxes on restaurants and hotels are progressive, since the rich are likely to spend proportionately more on these items, and small establishments and itinerant vendors are exempt. The effects of the registration taxes levied on businesses are uncertain, depending on who the final consumers are and what proportion of the tax is passed on. Finally, the tax on

non-motorised vehicles can be regarded as regressive, since users of bicycles and becaks are likely to be among the poorer members of the community. 7. Obstacles The actual process of implementation and the obstacles confronted thereunder raises issues which do not always directly impact upon human development, yet nonetheless underlie the whole discussion of decentralisation. This survey would be incomplete without an appraisal of the operation and significance of these obstacles. Indeed this is regarded by many commentators as the most important issue to address (e.g. Rondinelli, 1984). It will have emerged from the foregoing that decentralised systems of government should be appraised in terms of actual operation, rather than ostensible aims and designs. This is especially relevant in deeming particular decentralisation efforts as representing either deconcentration or devolution. Significant practical problems can inhibit the implementation of government efforts to decentralise. They range from the political - central decision makers are likely to resist any diminution of their powers - to the practical, especially shortages of resources and the difficulties arising due to weak local administrative capacity. Ministers of finance in the national government may oppose the transfer of revenue sources to decentralised units, whilst reliance on central funding can mean chronic delays, revenue shortfalls and reduced local autonomy. For example, it has been reported that the bureaucratic obstacles to decentralisation in Nepal were so serious that local officials were frequently required to travel to Kathmandu in order to expedite the release of central grants (Bienen et al 1990). Political constraints "Where decentralisation is proposed seriously, it is likely to be a focus for political confrontation, precisely because it has to do with specifying who may participate legitimately in decision making and whose interests will have highest priority" (Samoff 1990, p.524). It is predictable that change will be resisted by those who perceive their vested interests threatened by a new structure of governmental arrangements. This, coupled with bureaucratic inertia, can constitute formidable obstacles to the implementation of decentralisation reforms. Political will at the highest level may not, per se, be sufficient to overcome such obstacles. A number of authors criticise the assumption that national governments are "unitary, neutral actors". McGinn and Street (1986) argue that policy makers evaluate organisational reforms in terms of the impact upon their own power and control over the use of public resources - so that different actors react differently to the same policy.

Since 1983 the Rawlings government in Ghana has been pursuing a decentralisation proposal that would transfer virtually all ministerial activities

through the regions to the localities. In 1988, District Assemblies were elected which were supposed to take command of district-level administrations onto which the resources and expertise of central ministries were to be devolved. Olowu (1990) reported that progress has been very slow because of powerful forces within the administration against the plan; the distrust of elected councillors by military and civil servants alike; and the widespread feeling that decentralisation was a diversion from national economic policies. In the meantime, the old structures have continued through selected management committees presiding over deconcentrated administration at regional, district and local levels. An IRD project in Peru in the early 1980s which sought to combine regional spatial planning with decentralised administration was "ultimately undermined by political obstacles and central government controls" (Rondinelli and Wilson 1987 p. 348). The most serious constraints stemmed from continuing central controls on local planning and resource allocation. This "not only inhibited the local development corporations from responding to local needs and conditions, but subjected them to manipulation by the dominant national political party" (ibid p.353). The highly centralised control meant a vacuum of information about local conditions which "in turn generated apathy toward regional problems and this, combined with the lack of power or resources at the local level, created pervasive inertia against acting promptly on regional plans or local development proposals. In rare cases expeditious action occurred when local political groups bypassed departmental channels and went directly to the central ministries or influential congressmen in Lima. The study concluded that the potential for decentralisation depended crucially upon increasing political commitment to decentralised planning and decision making (ibid. p. 356).

Bureaucratic obstacles Beyond the political tensions involved in decentralisation attempts, there often arise significant problems related to the nature of the bureaucracy - its structure, technical competence and resource base. The creation of new systems of local government requires considerable administrative skill and resources. There may be basic problems with respect to personnel needs, attitudes (held both locally and centrally), and training in local government. In a number of cases, the central government has retained effective control over local personnel through recruitment, promotion, payment and so on. Although this may help to alleviate skill and resource shortages at the local level, it leads to accountability and loyalties which lead upward to the capital, rather than to local constituents. This may undermine local autonomy, and create an obstacle to the implementation of decentralising reforms.

In Indonesia, it has been observed that the structure and regulations governing the civil service severely handicap efforts to develop a local civil service adequate to meet the demands of decentralisation (King, 1988). Substantial deconcentration of

development activity through INPRES programmes was found to have been undermined by centralising civil service reforms (as well as financing arrangements, see ** above). In 1985 over three-quarters of employees working in regional governments were seconded central employees. Local staff available for decentralised management was subject to cumbersome and centralised control. Further, the structure of incentives encouraged a "brain drain" from the regions to Jakarta. In Bangladesh, although the services of civil servants has been formally transferred to the upazilas, their appointment, training, transfer, promotion and discipline all remain subject to national government control (Khan, 1987, op cit). In Botswana the recruitment, posting, training, promotion and discipline of local authority staff is generally the responsibility of the centrally directed Unified Local Government Service (ULGS). The ULGS staff at headquarters has been strongly criticised by the District Councils for not being sufficiently aware of, and responsive to, local problems and needs (Tordoff 1988 p.198).

Staff shortages at the local level appear to be a crucial constraint in many developing countries. An IMF survey conducted in 1982 showed that 57% of all government employment in industrial countries is at the 'local' (as opposed to the central) level, compared to only 15% for developing countries - Africa averaging only 6%, Latin America 21% and Asia 37% (World Bank 1983). This can be seen to reflect personnel shortages as well as the fact that local governments account for a greater share of public expenditure in industrial, relative to developing, countries.

In Nigeria, there is reportedly a wide gap between the staff required by local governments and the actual number employed by them. Local governments have not been able to attract and retain qualified staff and attempts by state and federal governments to train local government staff have not overcome this problem (Nwankwo 1984). Rondinelli et al (1983) emphasise the "crucial effects of shortages of trained staff on the success of decentralisation" (p.69). There are many examples of local governments with limited capacity. In the Chilean context of a relatively thoroughgoing decentralisation, the limited capacity of local officials has confined their work to mainly administering central directives: local initiatives have been rare (Racynski 1987). In Kenya the vast majority of skilled technicians and managers have been concentrated in Nairobi. The Sudan has suffered severe skill shortages, even within central ministries, which impeded decentralisation and limited the ability of provincial administrators to provide even basic services. In Pakistan, the technical officers available to the markaz councils were quite limited - in that the project managers are often either agricultural technicians or generalist administrators with little experience of district planning and development (Khan 1982).

The Moroccan government recognised the importance of personnel at the local level in creating a new special corps of posts in the local civil service, consisting

of centrally trained and paid chief civil servants called 'secretary-generals'. Training centres for local civil servants were greatly expanded (Nellis 1983). In Zambia, administrative capacity to plan and implement substantial development programmes was lacking at the "provincial, let alone district level (and) even in Lusaka". The shortage of qualified staff, the absence of suitable training programmes for district development officers, the unattractiveness of local government employment all contributed to this problem (Rakodi 1988). In Kenya there are several management problems at the local level. Many officers tend to be underqualified for their jobs; in some cases appointment is on the basis of political criteria rather than merit. There are frequently severe gaps in staffing. Some councils have six or seven vacancies at senior levels (clerk, treasurer, surveyor etc.) - so that one officer might be expected to fulfil all of these tasks over an extended period. In this light it is noted that in the 1984-88 National Plan period, even the more successful councils only managed to carry out about half the projects they had proposed (Wallis 1990). In Botswana the separation of the 'Tribal Administration' from the Village Development Committees is said to have stifled the effectiveness of the latter. Although the traditional institution has been stripped of its statutory powers it remains influential, as "the bulk of the rural people still look to the Chief and ... the traditional hierarchy for advice and guidance on a wide range of issues" (Tordoff 1988). Another example of institutional weaknesses undermining the efficacy of local government comes from Liberia. In the period 1974-1980, a World Bank-funded urban development project in the Liberian capital concentrated on infrastructural improvements. The Monrovia City Council had very limited spending and administrative authority. Local revenue covered only 7% of local expenditure, accounting systems were very weak, and mismanagement occurred frequently. The responsibilities of the MCC were poorly defined and qualified staff were lacking.

The World Bank worked outside of this structure through its "Project Team"; by the end of 1985 the planned civil works were largely completed, which benefited an estimated 130 000 people. However little had been accomplished in the way of institutional reform. The MCC became increasingly dependent upon central funds, as the government was becoming less reliable. One evaluation (by Price Waterhouse) identified the following weaknesses: (i) shortages of professional skills; (ii) responsiveness of the appointed mayor and council to their superiors, rather than local residents; (iii) inadequate revenue and weak fiscal discipline; (iv) uncoordinated development programmes; and (v) failure to enforce local regulations. The role of NGOs and traditional structures It can be argued that human development is such a complex issue that a wide variety of institutions - market, central government and field administration, local government and non governmental organisations - are required. Hyden, for example, writes in the African context that genuine local governments must build upon the wide variety of traditional

and modern local associations, because otherwise people will perceive local governments to be an extension of the State. Similarly, Uphoff argues that there should be "multiple channels of action". Since any one channel can become monopolised or blocked, it is said to be better that there be a degree of competition as well as complementarity between various institutions - whether bureaucratic, local government, co-operative, private enterprise, NGO, religious or other. More recently Wunsch and Olowu (1990) have argued along the same lines. The attention accorded to NGOs follows their apparent comparative advantage in areas of priority for human development, given their flexibility and small size, as well as highly motivated personnel (Paul, in Lewis 1988). However rather than attempt to assess the relative performance of NGO vis- -vis that of the various levels of government, the ambit of discussion is more limited. It investigates the proposition that local, rather than national, governments are better able to create an environment within which the work of NGOs is facilitated. Turning then to evidence of collaborative relationship between government and grassroots organisations, the following points emerge. There is already considerable operational collaboration in both planning and delivery of public goods for human development - e.g. Amul Dairy in India and the Grameen Bank (Paul, op cit). In Indonesia, the national government provides the funds, supplies and technical expertise needed for family planning programmes, whilst community volunteers and village organisations mobilise demand and assist in the distribution of contraceptive supplies. An important related issue is the role of traditional institutions. This has been explored in some detail in the African context. Traditional institutions have generally not been invested with measures of self-governance. In East Africa - Ethiopia, Kenya and Tanzania - and elsewhere governments have not treated them as critical agents in ensuring the appropriateness and sustainability of development initiatives (Harbeson 1990). Sometimes this has been due to their association with the former colonial regime, or with conservative obstacles to reform in the country. A number of writers have emphasised that national governments have neglected and even deliberately undermined traditional self-governing bodies; and that this, in turn, has inhibited the effectiveness of local government. "Rather than build and improve on these institutions, they replaced them with expanded state structures, often with disastrous results. Furthermore, historical and anthropological studies indicate that local self-governing structures are more compatible with several traditional African political systems than the nominally `decentralised' structures created by African states. These traditional self-governing structures (ethnic, town and community unions) have continued to co-exist with the newly created structures in several African countries with very little interaction between them" (Olowu 1990).

In Africa traditional institutions were the mechanisms by which the people regulated their social and economic affairs, and exercised and controlled political power. Traditional authority encompassed chiefs and councils of elders and the complex bodies of law which guided their actions; for example, the village councils promulgated and enforced access rules that regulated the balance between livestock, water and forage, and minimised overgrazing, and complex rules of access and use assured (young men across) most of Africa access to land. Since independence, the tendency has been to abolish any legal authority which remained in traditional institutions. Traditional leaders have, as in the colonial era, frequently been 'co-opted', through recruitment into the civil service or pensions. In some cases resistance to such policies has led to bloody violence - as in Southern Sudan, Uganda and Ghana. In the result, the absence of effective local institutions is said to inter alia threaten common resources and community endeavours. Deforestation and decertification processes in the Sahel have been traced directly to the absence of local institutions able to regulate tree use and livestock foraging. In Botswana, the abolition of local traditional authorities made it far more difficult to solve such problems (Wunsch 1990). Also, as noted above, traditional leaders appear to have been far more effective than government officials in collecting local taxation dues.

Occasional Paper 13 - DECENTRALISATION: A SURVEY OF LITERATURE FROM A HUMAN DEVELOPMENT PERSPECTIVE

PART III - SECTORAL SPECIFIC CONSIDERATIONS 1. 2. 3. 4. 5. Education Health Water and sanitation Housing Economic infrastructure

PART III - SECTORAL SPECIFIC CONSIDERATIONS The previous part explored such aspects of decentralisation as efficiency, equity and participation, supported by evidence drawn from the sectors most relevant to human development. This part seeks to portray a better sense of how decentralisation has tended to proceed in each sector, highlighting the most important characteristics. Thus, for example, the specific issues raised by decentralisation in education include the nature of decision making structures and interested constituencies, political ramifications and financing; whereas in the health context, decentralisation is very much associated with the concept of primary health care.

1. Education According to Winkler (1989), a highly centralised education system is perceived to be relatively expensive and inefficient, poor in quality, and unresponsive to the interests and needs of parents and local employers. Much of the evidence pertaining to the education sector was discussed above, under the general headings of Efficiency, Equity, Obstacles, and so on. The discussion in this section is confined to those aspects of decentralisation which are specifically relevant to education, in particular the nature of the decision making structures and issues of financing. 1.1 The degree and type of decentralisation In practice, there is great diversity in the extent and type of decentralisation of educational functions. Decisions about education are typically made at several levels national, state, regional, local, school, household and individual. At the same time, the system of education in any country consists of a number of components - ranging across curriculum and supervision, teacher recruitment and remuneration, and school construction. The degree of decentralisation is likely to vary according to each component. The organisation of schooling - minimum requirements and the structure of primary-secondary schooling - is often highly centralised. The chief administrator of a school district may be centrally appointed or subject to the local community; s/he may have substantial decision making authority, or be relatively powerless. Overall then, in order to assess whether a country's education system is decentralised, it is necessary to look at the distribution of decision making authority with respect to various educational functions. Most significant among these are: (i) The chief administrative officer -who selects the officer, and her/his decision making powers; (ii) Teachers recruitment, posting, payment, etc; and (iii) Local financial autonomy - the degree to which local authorities are able to finance their own education systems.

During the Cultural Revolution, the entire education system in China was decentralised, in the sense that national curricula and teaching materials were abandoned, as was the national examination system. Length of schooling, for example, was determined at the lower levels. It is said that the decollectivisation of agriculture and the decline of the communes and brigades since the late 1970s have deprived education of its major support structure. In the mid 1980s it was argued that the Chinese government's education policy was bifurcated - involving a small, centrally financed elite sector based on the "keypoint" concept, which trained first class scientists and engineers; alongside a mass sector providing basic education to the majority, which relied upon local government, collective and individual funding. The result is said to have been a serious neglect of basic education (Rosen, 1985). UNICEF (1990) reports that parents and village leaders have been actively involved in establishing and running BRAC schools. The village is responsible for providing the actual classroom (for which BRAC pays a small rent), and for deciding upon school hours. The monthly parents' meetings are well-attended.

The teaching staff, drawn mainly from the better educated women in the village, take part in an intensive 12-day training course and receive continuing guidance, support and refresher training. They are only paid a small monthly stipend. In Nicaragua after 1979, each school had an Advisory Council - involving students, parents and teachers - which were reported to be "committed to discuss educational policies in order to diversify curriculum, programme content and methodologies" (Cassusus 1990 p.10). There was, however, a tension between people deciding their own goals and course of action, as against the role of 'the vanguard', who were supposed to guide the process. Often the mass organisations, which were supposed to be mobilising the population around the tasks of the revolution - like the mass literacy crusade - were seen as little more than political forums for transmitting the political line being propounded by the FSLN (Arngrove and Drewes 1991 p. 102).

1.2 Financing The national government may both finance and directly provide all inputs into the school system, or the local community may fulfil part or all of this role. The system of financing recurrent local expenditure may be mixed - as in India, where the midday meal programme is provided by the local community but partly financed by central grants. The degree of local financing of education varies widely. In Brazil, for example, one fourth of all primary school funding is provided by the municipal governments, which is much higher than either Nigeria, where the local authorities run schools but provide little of the finance, or India where the central and state governments provide the vast majority of funds for education, and the zilla parishads relatively little (table 16) (Hinchcliffe 1989). In practice, school construction and financing tends to be among the most decentralised aspects of the education system, especially in Africa and Asia (Winkler 1989). In many countries the national government offers a matching grant, whereby the school constructed by the local community is staffed through central funds (as in Eastern Nigeria, Kenya, Zimbabwe and India).

The central government finances 64% of total education costs in China, while local governments, brigades, enterprises and so forth contribute 28% and "the masses" 8%. Over 90% of central funding goes to recurrent costs. At the basic levels education in China is not compulsory, nor is it free. In the countryside, primary schools and teaching staff are largely financed by the local population. In the mid-1980s, 80% of the funds used to build and repair middle and primary schools in rural areas were raised locally. Teachers are expected to raise funds through commercial activities (Rosen, 1985). The Indian government has played an increasingly active role in education, through (i) its own regional colleges, national scholarships, University Grants Commission etc; (ii) the centrally sponsored sector, where the states may undertake responsibility for implementation (e.g. promotion of Hindi in nonHindi speaking states); and (iii) the centrally assisted sector, where central interest is embodied in state plans, and its financial contribution ranges between 25-100%

of total cost. Direct federal education expenditures are limited however - there is a reliance upon intermediaries such as state governments and the UGS. In 1976, the reality was constitutionally formalised when education was put onto the concurrent list (it had previously been a state function). Still education represents a much larger proportion of state expenditure (about 20%) compared to the central budget - where education has declined from 2.3% (1976-7), to 1.5% (1985-6) total expenditure. Table * shows the centre-state shares in educational finances, by level of education (Tilak p.466). There have been calls to reform intergovernmental relations in education - Tilak (1989) recommends less physical concurrency (intervention in policy formulation, planning and administration), and more financial concurrency (devolution of larger resources to the states, or the centre helping the states to widen their own revenue bases). In some cases central grants are tied to specific educational purposes. In Nigeria, for example, it was estimated that 70% of federal and state grants to local governments support the national Universal Primary Education (UPE) scheme (Smith, 1982). In Zimbabwe, central grants for tuition must be used only for that purpose, as narrowly defined by the centre, and not for, e.g., school administrative expenses. 1.3 Political Aspects Decentralisation in education can be viewed in terms of the political implications and obstacles, especially vis--vis different groups within the bureaucracy, and national teacher unions and the State. In the Latin American context, for example, McGinn and Street (1986) stress that decentralisation is primarily a question of the distribution of power among various groups in the society.

Decentralisation in Mexico has been seen as a battle between the bureaucrats and the technocrats, both of whom had been incorporated into the ruling PRI. The latter were described as 'intellectuals with links with the private sector' - this faction was given a new Vice Ministry of Educational Planning in the Secretariat of Public Education, and the position of Education Minister under the Portillo government (1976-82), and sought to replace patronage in appointments with technical criteria. Part of the technocrats' campaign was said to lie in the decentralisation of decision making to delegates of the Minister, with authority to act in his name; followed by the creation of formal offices that transferred responsibility for all levels of education (except tertiary) to delegates, undermining the power of the bureaucrats in Mexico City. The National Director General of Education lost a major source of power when administrative control over teaching positions was transferred. The reforms were also intended to weaken the national teachers' union (an ally of the bureaucrats), by forcing it to negotiate separately with the 31 states. Ultimately however a dissident teachers' group which became a mass movement opposed to the union's central authority and antagonistic to both factions, eventually pushed the technocrats 'back into the arms of the bureaucrats', at least temporarily. This led to renegotiation of the decentralisation plan. The authors conclude from this and other case studies that

centralisation-decentralisation are chosen not only on their technical merits, but also as part of a strategy where political effectiveness is an important consideration (McGinn and Street 1986). Cassusus (1990) also emphasises the 'tensional power relationships' generated by the process of decentralisation in reference to Bolivian education. The teachers' union regarded decentralisation initiatives as an attempt to break their own hegemony, which prevented implementation of the project, at least during the Movimiento Nacionalista Revolucionario (MNR) government. In Thailand, non-formal education (NFE) has been deconcentrated following the recognition that the NFE curriculum ought not be urban-oriented in a largely (80%) rural country, and that there was a need for research and curriculum development outside the centre in order to overcome excessive standardisation. Regional centres of the Adult Education Division (AED) were established to conduct research into the educational needs, and produce NFE curricula and learning materials relevant to the ethnic and occupational structure of the region. The changes only affected the control of research and development activities since operations had always been a provincial responsibility. Armstrong (1984) found that implementation of these reforms was facilitated by the following factors: (i) Senior officials in Bangkok did not view NFE as a high priority activity (compared to, say, secondary education) and therefore did not interfere in the process. (ii) National policy had endorsed decentralisation as a (rhetorical) goal, whilst the relevant policy documents and guidelines were extremely vague, which facilitated a flexible approach at the local level. (iii) Decentralisation policy was formulated within AED, and key officials were enthusiastic about the project. In several cases regional directors were able to mobilise local resources and recruit highly qualified staff.

The study identified the most significant constraints to decentralisation as: (i) the highly centralised Thai budgetary system. Bangkok staff took an active role in the formulation of provincial budgets, thereby imposing their own programme decisions, whilst many provincial directors did not become seriously involved in the budget preparation process. A more significant inhibition was the cumbersome disbursement procedure; delays of up to four months into the fiscal year were common (not only for AED but throughout the system). (ii) The emphasis of the Thai bureaucracy on quantitative expansion. Since enrolment increases were most easily achieved in the existing urban-oriented programmes, it was difficult to move into the rural areas. This also resulted in a preoccupation with constructing and staffing new offices, rather than with evaluation or training. 2. Health A number of international organisations and academic writings have endorsed the decentralisation of health services. Centralisation is associated with an emphasis on high cost curative medicine, highly trained medical personnel and technologically advanced facilities largely confined to urban centres and the better off. Decentralisation is seen as a means of achieving greater responsiveness to local needs through the delegation of

authority and resources to the local community and to intermediate levels (WHO 1980). It has been praised as a means promoting local participation in health care (Mills et al. 1987). Community participation, in turn, has been "almost universally accepted as the cornerstone of primary health care in developing countries" (Asthana 1989). Access to basic health services is available to fewer than three in ten people living in the rural areas in a number of developing countries. The argument for decentralisation rests partly on the need to increase access to health care, particularly in rural areas. Evidence suggests, for example, that the spatial deconcentration of maternity care could significantly improve the serious rates of maternity mortality presently suffered in many developing countries.

One longitudinal study from Cote d'Ivoire found that the presence of an obstetrician in the district greatly reduced the incidence of complications (which had formerly necessitated transfers to hospital). The maternal mortality rate for rural women was 2013 per 100 000, compared to 33 per 100 00 in the town where the medical facilities were located. Others are more wary of attempts to promote decentralisation in the health sector, suspecting an association with the "contemporary international pressures on developing countries (which) militate strongly against a positive central government role in health", and seeing it as "a discrete way of the state abandoning its functions" (Collins 1989 p.169). There is a concern that "overdecentralisation" could break up the consistency and coherence of national health policy and constitute a major obstacle to improved equity (ibid p.170).

In what has been described as "the most radical decentralisation policy to date" (Winkler 1989), responsibility for the provision of basic social services in Chile was transferred from the State to local entities. The health care reforms mirrored the approach adopted throughout the social sectors, as: (i) responsibility for primary health care, infrastructure and personnel, was transferred to the municipalities; (ii) the role of the central Ministry of Health was limited to policymaking and supervision; and (iii) providers of health service, both public and private, were paid directly on a fee-for-service basis.

The impact of these reforms is controversial. In particular the redistributive implications (see Equity, above) and the increased central control (see Participation, above).

It is also interesting to note that in Europe, where it is "correct to say that the provision of health care has been decentralised, through the development of primary health care services ... there is still a large gap between the health-for-all strategy and what goes on in PHC services" (Godhino 1990). For instance, communication and collaboration between communities and hospitals, health centres and local PHC programmes "remains weak" (ibid p.47).

2.1 Implementation

A survey of the literature on health policy suggests that the district is increasingly regarded as "the backbone" of primary health care since, under the guise of various names, it is "the most peripheral fully organised unit of government" (Tarimo and Fowkes 1989). Certain advantages have been put forth: that it is administratively capable of supporting a specialised technical and managerial staff and exercising substantial decision making power, yet geographically and demographically compact enough to facilitate efficient management. Integration is put forward as an important principle to be pursued at the district level - e.g. collaboration and sharing of facilities between programmes for maternal and child health and immunisation. Also inter-sectoral concerns can be encompassed - e.g. the improvement of local water supplies. Yet while considerable decentralisation of authority is called for, it is often stressed that "overall national guidance and monitoring have to be provided by the government" (ibid p.79).

Some of the problems which can arise in the course of attempts to decentralise the control of health services were evidenced in Papua New Guinea in the early 1980s (Thomason 1984). The central Department of Health decentralised the management of health services concurrently with the commencement of a major rural health development project in several provinces. The main components of the project (e.g. constructing and equipping new primary health facilities) included the activities whose control had ostensibly been decentralised to the provinces. A special central government unit was responsible for overall management of project implementation, including day-to-day operations. Throughout the planning and development of the project, there was only limited involvement of the provincial divisions of health, and even less of the respective provincial governments. This was despite the fact that the project was affecting purely provincial functions towards which the provincial governments would be required to provide a substantial financial contribution. These arrangements lead to frequent misunderstandings and conflict between the relevant institutions, which was reflected in long delays in implementation. A more recent study of Papua New Guinea, concluded that the provinces are now effectively responsible for the day-to-day operations of primary health care facilities and programmes. Provincial governments are provided with central grants, to be spent at their own discretion to organise and administer health programmes. Many provinces have decentralised even further by creating district offices designed to co-ordinate the local activities of the specialised agencies. In some cases district health officers are appointed, who are responsible for the supervision of health workers, submission of the district health budget, and making such policy decisions as where to locate new facilities. Generally, however, the district structures did not significantly further decentralisation budgets were determined by the province and the district input was basically advisory. The study concluded that the provincial level was the most appropriate for responsibility for health care, combining support and supervision with local flexibility. In Mozambique, there was a shift in emphasis from hospital- based curative care to preventive medicine, combined with a broader infrastructure of basic curative care administered by intermediate health workers. The government reduced the

list of essential imported drugs to 300, purchased on a competitive tender basis so that the proportion of the health budget spent on drugs was maintained within a range of 10-20% (compared to 30-40% for most developing countries). These shifts met considerable opposition from health workers on the grounds of quality. Attempts to introduce courses on public health in the universities met with equally strong opposition from medical doctors and students alike. There was also local demand for basic curative care (see above). Further pressures were encountered by the new health policy in Mozambique. Real health expenditures fell in the face of escalating war costs and economic crisis. Donors increasingly by-passed the state drugs importer and failed to respect the national drug code. There were also organisational problems as aid agencies tended to favour selective health programmes (e.g. for immunisation and diarrhoea disease control) each with their own separate funding, wherein key decisions were typically taken by a small group of foreign experts. This approach was said to have left little room for consultation with, or the involvement of, the local people in villages. Mackintosh and Whyt (1988) argue that in Mozambique, health policies had been made nationally, but with genuine pressure from below as a result of decentralisation to local communities, but that donor practice effectively undermined this system (pp.158-9).

A USAID-assisted project in Indonesia was partially effective in promoting decentralisation of health management, planning and implementation (Bossert et al. 1991). The capability of lower units was enhanced vis--vis the highly centralised Ministry of Health, particularly in the collection and analysis of better data (in terms of quality and quantity) than was available at the national level. Provincial and local health officials could convince central officials that national policies should be modified in order to take account of provincial realities. They were able to challenge national targets (especially those for specific diseases), initiate specific drug proposals that were accepted by the centre, and implement changes in national policies that were innovative or more appropriate to local needs (such as a TB programme in a Kabupaten in West Sumatra). Lower level units were able to persuade Jakarta officials that their decisions about priorities and interventions were more appropriate for the province than those set nationally. Another dimension of the project lay in innovative training of nurses, which encouraged greater community involvement and longer periods in the field. The result was not, however, thoroughgoing devolution of decision-making in the health sector. Often local initiatives were modified by the centre, rendering them 'less effective'. Local responsiveness to the decentralisation project was not always positive - many local and provincial officials were not willing to support additional health efforts, and those who were seldom committed significant resources to support project initiatives. The study concludes that control of information is a necessary, but not sufficient, condition for effective decentralisation of decision making.

3. Water and sanitation

Here, perhaps more than any other sector, in schemes for the provision of water and sanitation, the literature emphasises the importance of community involvement. This involvement should encompass all stages, from project initiation and design to construction, operation and maintenance. 3.1 Decision making processes It is generally held that centrally directed engineers, with little reference to local needs, are unlikely to realise the full potential of schemes for the provision of water and sanitation. The lessons which have emerged from experience emphasise the deficiencies of a top-down strategies. Local people should take a leading role in planning, constructing, financing and managing rural water supply and sanitation projects (IBRD 1987, 1988; Narayan-Parker 1988; Douglas 1990; U Tin U 1988).

It is reported from Adayaw in Burma that the network of basic health services staff and volunteers plays a vital role in encouraging community participation in development projects. Under the supervision and guidance of the local party and council, the town launched a major water supply effort, resulting in 97% of the population having access to safe water (U Tin U 1988). Nonetheless, government - local or central - has an important supporting role in this context.

The Malawi Self Help Water Scheme, which is rightly regarded as the outstanding water project in rural Africa' (IBRD 1988), reflects the success of community collaboration with the national government. (The scheme was described on p.16 above.) Community participation has been successfully "institutionalised" in water committees (Cox and Annis 1982). The committee can seek technical assistance, channel requests for water, organise and take responsibility for the mobilisation of local resources, draw up work schedules, and promote community awareness about health education and hygiene.

It is especially important to obtain the participation of women, who bear the burden of water collection, yet are often excluded from community decisions. In most villages, women and children are also the primary victims of water-related disease (Cox and Annis 1982). When systematic attempts are made to include women in project development and maintenance, the results have been "encouraging" (IBRD 1988). Opportunities can be provided to enable women to improve their technical skills to perform these tasks. Examples from Zambia, Bangladesh, India and Paraguay reflect the training and involvement of women in maintenance. Yet in Malawi, where women provide most of the self help construction labour - they constitute only 10% of the important planning committees. At the same time it is important that such projects do not increase the already heavy workload of poor women, especially when their participation is voluntary rather than paid (Yudelman 1986). 4. Housing

A general lack of data on housing availability, quality, tenure and so on makes any assessment of housing conditions in developing countries difficult, if not 'conjectural' (UN 1989). Nonetheless it is clear that the housing available to most households in developing countries - especially to the poor - is grossly inadequate (Struyk 1988). This problem is especially acute in rapidly growing urban areas - squatter settlements, shanty towns, and low-income neighbourhoods (Annis 1987). Publicly supported contractorbuilt units, hitherto the norm, have not been able to meet this deficit - especially in cities growing at an annual rate of 3-5%. Since the mid 1970s(?), there has been a shift in public policy. The role of the state in housing policy is increasingly seen as that of an enabler, rather than as a provider. The focus has shifted to sites and services, slum upgrading, security of tenure and the provision of basic infrastructure. The issue here revolves around the extent to which decentralisation can help alleviate the problem of providing decent housing, especially for the poor. There is a considerable body of literature which appraises the role of community groups and self-help here. We also need to compare the role of local, to that of the national government, especially in the sense of its activities as an enabling institution. 4.1 Role of the government The activities of governments, national and local, have a bearing upon the housing sector in the following senses. Various public policies may facilitate or deter informal housing investments, such as credit policies, building regulations, the nature and security of land tenure, and the regulations which affect the use of residential plots for income generating activities. The provision of basic infrastructure - water, sewerage, roads, electricity - also affects private housing decisions and the quality of housing. Some activities are likely to fall largely into the central domain - including real property legislation, credit policies and the provision of, say, electricity. In the remaining areas, there may be important steps open to local authorities to contribute to an 'enabling environment' to facilitate housing development. Local government is arguably in a good position to pursue such strategies. It can engage in consultations with local people, better appraise local conditions and needs, and frame policies and regulations appropriately in that light. There are however cases where municipalities have acted in ways which have not enhanced the availability of decent housing for the poor. Throughout Latin America, there have been numerous land invasions, with groups of low income families occupying undeveloped plots of private or public land and erecting makeshift shelters. Ironically in Colombia, much of that land was purchased by municipalities for use as sites as lowincome housing, which had yet to be built (Sorock 1984). In some instances the municipal regulations that were designed to protect homebuyers by requiring private developers to build adequate infrastructures have aggravated housing shortages (e.g. in Colombia).

In Iloilo City in the Philippines, the city government is said to have failed to respond to housing needs, which is one of the most serious problems facing the city. This was partly due to a lack of funds, and also to a "generally negative attitude towards squatters on the part of decision makers" (Ruland and Sajo 1988). When asked about the most effective strategy to approach the squatter problem, almost half the councillors opted for 'resettlement to prepared lots in periphery areas', although many of them were well aware of the adverse effects of resettlement on the living conditions of the families (ibid p.273). In Colombia, self-help housing movements (groups) receive technical assistance from SENA (the national institute for vocational training), and municipal planning authorities assist groups in designing site and infrastructure plans. Municipal officials also provide technical assistance to help upgrade local homes and communities. Municipal public utilities assist the groups by drawing up plans and budgets for the water, sewerage, and electricity systems and by supervising the community workers on the site (Solock 1984). Self help housing groups in Argentina joined together to pool expertise in architectural, financial and social services, and also to increase overall bargaining power. In the 1983 election campaign, the confederation urged the legalisation of land ownership where necessary, and support to self-help housing programmes through government credit subsidies, better technical assistance and less red tape (e.g. simpler building codes). They called on the state to provide basic infrastructure and public services. Nonetheless it was accepted that the national economic situation prohibited the construction of large public housing projects or increased housing subsidies, and that improvements must be sought primarily through legal and policy changes to facilitate self-help programmes (Page 1984).

The self-help movement does not lack critics however. It has been noted that propertyowners and landlords rather than tenants, and older rather than more recent residents, tend to dominate residents' committees in redevelopment and upgrading projects (Stren, in EDI 1989). 5. Economic infrastructure An array of economic infrastructure - including roads, irrigation, agricultural extension services, basic communications and electricity - facilitate local economic and employment growth. There is a growing body of opinion which regards the efforts of the central government in this regard as manifestly deficient (see Wunsch 1990). Attention has thus focused upon the role of the community and local government in the creation and maintenance of economic infrastructure, particularly in rural areas. At the same time it is recognised that national macro-economic policies, regarding fiscal and monetary matters, pricing and sectoral policies etc., will generally have a crucial impact upon economic activity throughout the country.

One detailed micro-level study of a cluster of villages in West Bengal focusing on ways to promote employment expansion, emphasised the role of the 'panchayats' the local unit of self government. The sectors covered included agriculture,

animal husbandry, food processing, cottage industries, transport, and the building and maintenance of rural infrastructure. It was noted that "these works will involve a considerable amount of micro-level design, planning and organisation of human-power and materials, which cannot be achieved by the traditional bureaucratic methods of administration". The author concludes that local popular institutions of self-government, with appropriate leadership, can be expected to be much more useful in promoting employment expansion (Maitra 1982). Sanwal (1987) reports from Uttar Pradesh, India that decentralisation increased the commitment of local officials to promoting local development. Block Development Officers took bank managers out with them to the villages, the banks fixed dates for groups of villages in order to expedite loan procedures, and villagers were taken in groups to markets to select the best stock, carts and implements.

5.1 Decision making procedures The responsibility for making decisions about the provision of economic infrastructure can be decentralised in a variety of ways. The following paragraphs present some country experiences in point.

In the Indian state of Karnataka, legislative reforms in 1987 brought the planning and management of a number of important development activities within the jurisdiction of the zilla parishad (district). The zilla parishad is responsible for extension and training, the operation of commercial and seed farms, the development of irrigation, and the promotion of small scale industry through the establishment of training, production centres and marketing facilities. The most novel aspect of these arrangements lies in the legislative intention to give the zilla parishad effective control over the line departments. In the past, local authorities had exercised only limited control over the spending departments within their jurisdiction. According to Slater and Watson (1989) however, it seems likely that what public investment is planned will remain subject to national or state level policy decisions translated into targets, whilst how much investment is planned depends upon financial allocation procedures - decisions which are generally made in the state capital. The authors concluded that the main scope for manoeuvre lies in where investment is planned - which arguably had always been subject to local control. The operations of local government in Zimbabwe's Communal Lands in economic infrastructure are very restricted in scope - district councils spend very little on capital investment, as much of the investment on infrastructure (roads, water etc.) is made by the central government through the District Development Fund (Mutizwa-Mangiza, 1990).

Occasional Paper 13 - DECENTRALISATION: A SURVEY OF LITERATURE FROM A HUMAN DEVELOPMENT PERSPECTIVE

PART IV - EXPENDITURE ALLOCATION - COMPARATIVE PRIORITIES

1. 2. 3. 4.

Introduction and summary Local community priorities Local government priorities Some Empirical Evidence Bibliography

PART IV EXPENDITURE ALLOCATION - COMPARATIVE PRIORITIES 1. Introduction and summary One task of the present study is to identify the overall relationship between decentralisation and expenditure allocation by sector - that is, whether or not local governments are associated with relatively higher expenditures on areas of priority for human development. Following the Human Development Reports of 1990 and 1991 (as well as a substantial body of academic work and numerous international declarations), "human priority" is defined to encompass those expenditures which contribute most to human development. Basic education, primary health care and family planning, water and sanitation for the poor and decent housing all fall within this ambit in the social sectors. A number of categories of economic expenditure, on rural roads, irrigation etc. are clearly also important, although excluded for the purpose of the present exercise. First this Part sums up the general arguments about expenditure allocations and decision making under decentralisation - in terms of the priorities of the local community and local governments. This is illustrated by a number of case studies. It is suggested that decentralisation has demonstrated potential to lead to a greater emphasis upon areas of priority for human development, but that this need not follow. Second, available evidence as to the relative priorities in specific sectors is presented. Finally, but certainly not lastly, an attempt is made to quantify the ratios of local and central governments. The limited evidence available reveals such a diversity of country experience that it is difficult to arrive at any firm general conclusions. Nonetheless it can be said that the sample revealed that proportionately higher shares of local expenditure are typically directed to the social sectors.

At the outset, the difficulties facing the present exercise should be noted. Any attempt to assess relative priorities is problematic where local government cannot be regarded as 'autonomous'. The vertical fiscal imbalances and lack of financial autonomy which characterises local government in many countries may well entail a lack of local competence to determine local expenditure allocations. Central grants may be directly tied, or subject to guidelines. Further it is possible that decentralisation may impact upon central expenditure allocations, leading to higher, or lesser, shares directed towards human priorities. Note should be made of the obvious point that local priorities are virtually inherently more likely to accord to the basic priorities of human development, in the following senses. First because local governments are not generally concerned with such matters as defence, general administration and overseas debt repayment which may consume a large proportion of national budgets. Second, within the social sectors the focus is likely to be upon the lower (cheaper) levels of provision with their primarily local effects - for example, local clinics and dispensaries rather than large sophisticated hospitals, and basic schooling rather than universities. When examining decision making under decentralisation there are at least two related dimensions - the priorities of the local government and those of the local community. The main focus here is upon the former given that the task of the survey is to examine decentralisation of government. The latter is, however, clearly relevant - indeed in systems characterised by a significant degree of participation in local government, the priorities would be expected to converge. 2. Local community priorities It is difficult to generalise about the priorities of local communities, which will inevitably vary across inter alia urban, rural and remote, better off and poorer, industrial and agricultural areas. Nonetheless the following observations can be made. The priorities of the local community may well differ from those considered to be important by the national government or donors. One major study argued that there is good evidence that neither rural nor urban residents volunteer for communal projects aimed at increasing productivity (e.g. irrigation); rather the projects in which they are prepared to participate showed that their priorities lie in the construction of schools, health clinics, water supplies and roads (Ralston et al 1983). It has been observed that the self-help activities are mainly concentrated upon the provision of priority social services such as schools, water supply and clinics (Hughes 1985 p.60). Insofar as self-help schemes originate within local communities, they may provide a forum for expressing local needs and priorities. (This is not, however, the case when they are dominated by influential but narrow interest groups; when the poor lack the skills, confidence or influence necessary to advance a case; or when the availability of matching funds distorts local demands.)

In Tanzania, rural communities involved in the self-help movement co-operated in building schools, dispensaries and water supplies, but did not participate in agricultural projects unless compelled to do so by the government (Hyden, 1968 cited in Ralston ibid.). The same pattern has been found in Zambia (projects of the Ward and Village Development committees), rural communities in Papua New Guinea and in the highlands of Peru (Ralston, ibid.). It has been found that women's groups in most developing countries tend to be focused on health and education, rather than physical infrastructure, such as roads (Miller, 1982, p.6-9). In 1985, in response to demands from village parents, the Bangladesh Rural Advancement Committee (BRAC) started experimental schools aimed at providing basic literacy, numeric and social awareness to the children of landless families. BRAB has developed a new model of basic primary education, which has proved to be accessible to the children of the poor and relevant to their needs. By late 1989 the programme had expanded to 2500 schools, with a further 2000 to open by this year. Nigerian rural communities are reported to regard improved water supplies as of the highest priority, but because this is expensive are more likely to begin with simpler, cheaper facilities such as roads (Okpala, 1980 cited in Ralston ibid.).

The health priorities of local residents may differ from those considered of prime importance by the HDR. More specifically, the demand for curative care may be significant, so that primary health services may not meet local community preferences. Moreover, the poor may (understandably) be more concerned with basic consumption needs, prior to primary health care.

An evaluation of an NGO-assisted programme to train village health workers in El Salvador reveals some of the difficulties which may be encountered (Capps and Crane, 1989). A survey conducted among the health promoters, who were all volunteers with limited education, found that much more time was being spent on treating illnesses than on community health education. This arose from the public demand for curative care, especially medications. In some cases this was due to a single promoter serving a large isolated community where no other resources existed. It was also noted that advertising has lead people to expect modern medicines to provide quick cures for almost all illnesses. In Mozambique, many of the village health workers met real difficulties as they could not satisfy the local demand for basic curative care. This sometimes resulted in the people losing confidence in the health workers, which had negative implications for the effectiveness of preventive measures, and also for the decentralisation of financing health care. Subsequently, the Ministry of Health decided to adjust its services to local aspirations, through providing greater curative skills in the training of health workers (MacIntosh and Whyt 1988). Tumwine (1989) provides a number of examples from the Chmanimani district of Zimbabwe where the priorities of local people differed from those of donors, and of local health workers. He cites the case of an externally assisted clinic, which had only involved the community in the construction of the building. It is the least

utilised health facility in the district, mainly because the local community comprises va-postori (apostolic) religious groups which do not believe in the use of Western health services. Another example is the case of a village meeting called by a local nurse in response to a high incidence of bilharzia cases. The local health team decided that the people should be motivated to build and use improved latrines. But at the meeting it was revealed that the problem regarded as most serious by the community was 'sadza' (the staple food) and 'pellagra' (a nutritional problem). The meeting ended up discussing drought relief, nutrition, and communal food production plots. Thus the poor may be more preoccupied with satisfying immediate consumption needs than with the provision of services such as education and health. Gilbert and Ward (1985) found in their study of several Latin American communities that the level of participation varied according to the needs perceived by the local community. The lack of water and electricity attracted a higher degree of participation than the paucity of health centres and schools. In Zimbabwe, local participation in village community meetings varies according to the agenda to be discussed, the season, weather, etc. Land resettlement proposals, for example, arouse a lot of interest (see case study). Generally it has been noted that local NGOs tend to engage in areas of prime significance for human development - income-generating (e.g. credit), communal (such as afforestation), and social (e.g. primary health care) (World Development 1987). An evident priority of many local NGOs appears to lie in the task of educating and mobilising the poor to participate in, or demand services when they are illiterate, unorganised and/or live in remote areas. 3. Local government priorities It would appear unwise to make sweeping statements about the extent to which local governments focus upon areas of significance for human development. Many complex factors will affect the orientation of local governments - including the socio-economic environment; the extent to which local public authorities represent the interests of their constituents; the political climate, which facilitates or restricts the free expression of popular feeling; and the functional responsibilities and financial resources available, or the extent to which local representatives are able to make decisions relating to the planning and implementation of human development projects and programmes. Nonetheless this sections attempts to draw a general picture of local government priorities. Local governments often have significant expenditure responsibilities in the areas of basic social services. Generally their concerns will tend to vary between localities, particularly between rural and urban areas. Matters such as refuse collection, sanitation, street cleaning and zoning - residential and industrial - are naturally more likely to be addressed by urban councils.

One study of alternative forms of popular participation in the context of basic needs concluded that local representative institutions may be well placed to articulate local perceptions and priorities but "in practice, lack of information, sectional interests, lack of resources and/or restricted powers may prevent them from fully appreciating the needs of deprived groups or acting to remedy them" (Hughes 1985). It has been argued that political and administrative decision-makers, particularly in democratic systems, prefer short term projects with immediate returns (Bahl et al 1984). This problem is not peculiar to developing countries. There is said to be a bias of local government in favour of highly visible projects such as new municipal buildings, rather than infrastructure development and recurrent expenditures such as road repair and school maintenance.

It is reported that 80% of block grants to the upazilas in Bangladesh are spent on high profile and prestigious physical infrastructure, whilst allocations for critical social programmes are under-spent. Thomas (1989) concluded that, "without a further change in local level priorities, any further decentralisation of responsibility for primary education, health, water supply and other basic services to the local level cannot be contemplated" (p15). By way of contrast it was found in Agra district, Uttar Pradesh decentralisation promoted human development priorities: " The allocations for sectors like agriculture, social welfare and education...increased substantially ... Officers did more touring, and so came into direct contact with the poor, rather than dealing only through intermediaries. Programmes that benefited the poor and neglected areas were given priority" (Sanwal, 1987). Olowu (1989) reports that the experiments in self-governance across Africa in the early 1950s reflect high expenditures on areas of priority for human development, and that they had "remarkable success at building basic infrastructure" - roads, schools, clinics, bridges, markets, etc. - some of which still exist today (see table 18).

In the following sub-sections, examples of comparative priorities are drawn for specific sectors - education, health, water and sanitation, and economic infrastructure. (The extent of treatment of each reflects the availability of evidence, rather than the relative significance of the sector.) 3.1 Education In principle, it would appear likely that local governments would focus more upon primary and lower secondary levels of schooling, rather than tertiary education - given the largely localised effects and lower costs involved. However whether education, per se, is an important priority of local governments is another matter. This would depend, to some extent, upon the allocation of functional responsibilities among different levels of government.

There is little empirical evidence as to whether decentralisation will help to promote spending on basic education, in terms of comparative expenditure ratios for local and central government. At the same time however there are numerous cases of local initiatives resulting in the expansion of provision of basic schooling - BRAB and Harambee are notable examples which have been referred to elsewhere.

It has been reported from Uttar Pradesh, India, that decentralisation resulted in vastly increased provision of primary schools. There was increased emphasis on primary education, for example - in 1982, 98 schools were opened in Agra district, while the previous average had been seven new schools per year (Sanwal, 1987). The Chinese government under Deng endorsed increased spending on education in order to promote economic development. The centre expected lower level governments to use the increased revenue available under economic decentralisation to support basic and vocational education. Problems have arisen however (Rosen, 1984). The localities, who are supposed to bear the burden of financing of mass education, reportedly prefer to invest in heavy industry projects, which are more prestigious and compensate for the irrational and inefficient supply of national industrial projects. According to Rosen (1985), local officials perceive education as a form of consumption which is essentially a "bottomless pit". Since 1979, intergovernmental transfers from the centre have been made on a lump sum basis with broad guidelines, whereas previously education funds had been earmarked for very specific purposes. In the result, the mass rural sector is locally (under-) financed, compared to the wholly urban elite sector which is disproportionately and centrally financed (Sautman, 1991). (The foregoing may represent a good example of a standard type of principal-agent problem in the decentralisation context).

3.2 Health It is likely that local health expenditures are directed towards more basic levels of provision. In most cases, large, sophisticated hospitals would not fall within the fiscal possibilities faced by local government. This points to a higher share of priority health expenditures in local budgets. And at least in operational terms, it appears that certain areas of health care crucial to human development require a significant degree of decentralisation. Family planning, for example, may require medical personnel to go out to the people, and meetings with village leaders. The inherent limitation of population policies implemented through central agencies is expressly recognised in Bangladesh's current Fourth Five Year Plan (the national contraceptive prevalence rate is 35%). It has been argued that a shift in spending patterns alone will not be sufficient to promote preventive and basic curative health care - that there "also has to be a sharp change in the organisation of the social production of the new care" (Mackintosh and Whyt, 1988). An important difference between preventive health care and curative medicine is that the former requires the mobilisation of whole communities to be effective, while the latter normally treats individuals. In other words preventive care has to be eminently social in

character (hence, the term 'public health') (ibid, p.169). It follows that health policy, in order to be effective, requires a significant degree of devolution, and must draw on and interact with community organisations.

In 1972, the responsibility for health care in Tanzania was shifted to regional and district development directorates in a restructured central government field administration (Maro, 1990). The objective was to promote equity in the allocation of resources to health facilities. The establishment of rural health centres was funded partly through national government and partly through rural self-hElp, in the form of local materials and labour. Allocations for health increased greatly between 1972 and 1978, especially in recurrent expenditures, most of which went to rural clinics and dispensaries. This was accompanied by mass education campaigns aimed at improving nutrition, maternal and child health programmes. Maro (1990) found that most clinics and dispensaries were adequately maintained in terms of staff and essential drugs, and were generally regarded as adequate by users. In Mozambique, especially in the early years, local debate and as well as the decentralisation of resources were effectively promoted in the health sphere (Mackintosh and Whyts, 1988). The process of decentralisation in health planning and financing led to a shift in relative, and sometimes absolute terms, from city hospitals to rural clinics; from complex medical intervention to simpler procedures; from doctors to para-medicals; and from curative to preventive care (Macintosh and Wuyts,1988). There was an emphasis on training para-medical personnel and health posts. Health workers, chosen by local residents and sent on a six-month training course, were to concentrate on preventive care and health education, and be financially supported by the villages. Many villages put considerable local resources into providing health facilities and financing the health worker; some collected contributions in cash or kind from each household. There were, however, many problems with this system. The communal villages themselves were sometimes badly sited and unviable. This meant that health workers could not be financially supported, a fact which they naturally resented, given that teachers and agricultural workers were paid by the state. Also it was often difficult to maintain adequate basic supplies (Mackintosh and Whyts, 1988). The statutory responsibilities of the District Councils in Botswana include primary health care; the Councils are said to have played a major part in enabling the country to achieve one of the best health records in Africa (Tordoff 1988). Health services were not a high priority of local government in Ioilo City in the Philippines, where less than 6% of total local expenditures are directed toward health care. Health facilities suffer from inadequate premises, poorly qualified personnel, and shortages of medical and other supplies.

3.3 Water and Sanitation Extensive appraisals conducted in the course of the International Drinking Water Decade by UNDP, the World Bank, and others have endorsed, inter alia, the following propositions: (i) Safe, accessible water supplies are an important priority of people living

in rural areas in developing countries, as reflected in their willingness to pay for improved services and the extent of community involvement in a number of countries. (ii) Investments in sanitation services do not appear to be of high priority to most rural dwellers.

In Senegal, it was found that spending decisions under decentralisation generally responded to the priorities expressed by the local councillors - the largest share of rural community budgets are directed toward the development of water supplies, particularly wells. Data from the Kaolick and Fatick regions indicate that expenditure on water by rural communities averages between 30-40% of the annual budget, compared to health (7%) and education (14%) (Vengroff and Johnston, 1987). That clean water is a high priority for local communities is evidenced by the many examples of their successful involvement in the planning, implementation and management of water supply systems, including Burma (U Tin U 1988), China, Colombia and Malawi (Briscoe and Ferranti 1988).

In China the local community plays a leadership role in initiating water projects, selecting the source, technologies, and service levels, and financing the recurrent costs (local governments and entities are expected to pay capital costs) (ibid. p.19). Rural access to water stands at 66%, compared to a developing country average of 50% (HDR table 9).

4. Some Empirical Evidence Significant difficulties face any attempt to measure priority public expenditures in developing countries (for a brief, understated description, see HDR 1991 Technical Note 7). The prospect of quantifying local government priorities, and proceeding to conduct a comparative analysis, is even more daunting, given: (i) Severe data limitations - only 8 of the 85 developing countries which appear in the IMF GFS 1990 contain any data as to "Expenditure By Function" at local levels of government. Even here, we are limited to broad sectoral categories, without any further breakdown. Simply "education", for example, without any clue as to such sub-sectors as primary and secondary. (ii) Lack of financial autonomy - which may render any picture of "local expenditure allocations" illusory insofar as the local governments may not exercise any effective discretion in this sense. Even if central grants are not specifically tied, there may be guidelines which restrict the use of funds. Nonetheless, it is an important and useful exercise to undertake, at least in order to illustrate the general pattern of priority expenditures which exists. Table 19 shows the available evidence as to relative expenditures in the areas of education, health, and community amenities. The latter comprises public expenditures on the support of housing activities, community development, water and sanitation services, and social security and welfare services for the unemployed, the aged, and disabled people.

This data reflects an interesting picture of expenditure allocations. First as to education, in five of the eight countries, local spending is proportionately much higher than at the central level. Again in health, in five countries (although a different group from that for education), local spending is relatively higher. In six countries, local spending on community amenities was relatively more significant at the local level. It is also important to bear in mind the sub-sectoral breakdowns of expenditure. As the 1991 HDR emphasised, the need for restructuring within each of the social sectors is as urgent as the need to reallocate expenditures between sectors. In many developing countries more then 80%, and in some cases over 90% of central health expenditures are directed towards hospital care. Thus even where the central government spending on health is proportionately greater than that at the local level, the impact upon human development may be less. In Chile, 6% of central expenditure is directed towards health, compared to only 3.4% at the local level; but almost 96% of the central allocation goes to hospitals. Unfortunately, as noted above, sub-sectoral breakdowns are not available for local government spending. It might be surmised that local government spending would tend to be directed towards clinics and community health workers, rather than large expensive hospitals. [Arguably, however, such a bias ought not to be taken for granted, insofar as this would purport to answer an important aspect of the present investigation simply by assumption; if no general propositions can be put, and the matter would have to be approached on a case-by-case basis.] TABLE 19 Comparative priorities of local and central government: selected countries

Local government

Central government

Note:

LE refers to local public expenditure as a % of GNP - i.e. district, not provincial level of government CE central government expenditure as a % of GNP CA refers to Community Amenities (including social welfare) n/a Not available BIBLIOGRAPHY Anise, S. (1987) Reorganisation at the grassroots. Its origins and meaning. Grassroots Development, 11:2. Anise, S. (1988) What is not the same about the urban poor: the case of Mexico City, in Lewis (op cit.) Apkan, P. (1990) Local government structure as a spatial framework for rural development in Nigeria. Public Administration and Development, vol.10: 263-275. Armstrong, G. (1984) Implementing educational policy: decentralisation of non-formal education in Thailand. Comparative Education Review, 28: 454-66 Arngrove, R. and Dewees, A. (1991) Education and revolutionary transformation in Nicaragua, 1979-1990. Comparative Education Review, vol.35; no.1: 92-109. Aryeetey, E. (1990) Decentralisation for rural development: exogenous factors and semiautonomous programme units in Ghana. Community Development Journal, vol.25 no.3. Axline, W. (1984) Decentralisation, provincial policy making and the law in Papua New Guinea. Public Administration and Development vol 4: 305-23. Bahl, R. Miner, J. and Schroeder, L. (1984) Mobilising local resources in developing countries. Public Administration and Development. Bahl, R., Holland, D. and Linn, J. (1990) Mobilising Resources in Developing Countries, in Bird and Oldman (op cit.) Bahl, R. and Linn, J. (1991?) Urban public finances in developing countries. Oxford University Press (for the World Bank). Bell, T. (1987) International competition and industrial decentralisation in South Africa. World Development, vol.15; no. 10: 1291-1308. Bennet, R. .J. (1982) Central grants to local governments. Cambridge University Press, Cambridge U.K. Beradi, J., Richard, A., Djanhan, Y. and Papiernik, E. (1989) Decentralisation of maternity care. World Health Forum, vol. 10: 322-326.

Berman, P. (1989) Cost efficiency in primary health care: studies of health facilities in Indonesia. Health Policy and Planning, 4(4): 316-322. Biennen, H. et al (1990) Decentralisation in Nepal. World Development, vol 18: 61-75. Bird, R. (1990) Intergovernmental finance and local taxation in developing countries: some basic considerations for reformers. Public Administration and Development Bird, R. and Oldman, O. (eds.) (1990) Taxation in Developing Countries. Johns Hopkins University Press. Bird, R. and Miller, B. (1990) Taxes, User Charges and the Urban Poor, in Bird and Oldman. Bonney, N. (1982) Local government and political development in Papua New Guinea. Public Administration and Development, Vol.2: 113-127. Bossert, T. et al (1991) "Bottom-up" planning in Indonesia: decentralisation in the Ministry of Health. Health Policy and Planning, 6(1): 55-63. Brennan, G. and Buchanan, J. (1980) The power to tax. Cambridge University Press, Cambridge U.K. Briscoe, J. and deFerranti, D.(1988) Water for rural communities -helping people help themselves. IBRD: Washington DC. Briscoe, J. et al (1990) Towards equitable and sustainable rural water supplies: a contingent valuation study in Brazil. World Bank Economic Review, vol. 4; no.2: 115134. Bulpitt, J. G. (1972) Participation and local government: territorial democracy. in Parry (1972) op. cit. Campos-Outcalt, D. and Newbrander, W. (1989) Decentralisation of health services in Papua New Guinea. World Health Forum: 347. Capps, L. and Crane, P. (1989) Evaluation of a programme to train village health workers in El Salvador. Health Policy and Planning, 4(3): 239-243 Caputo, E. (1989) Enabling local economic growth, in EDI (1989). Casassus, J. (1990) Decentralisation and deconcentration of educational systems in Latin America: foundations and critical aspects. Major project in the field of education in LAC no 22: 7-16, World Bank, Washington D.C.

Cernea, M. (1988) Nongovernmental organisations and local development. World Bank Discussion Papers, no.40. Chikulo (1985) Reorganisation for local administration in Zambia Public administration and Development. Churchill, A. et al (1987) Rural Water supply and sanitation - time for a change. World Bank Discussion Paper 18: Washington DC. Cochrane, G. (1983) Policies for strengthening local government in developing countries. World Bank Staff Working Papers, no.582. Collins, C. (1988) Local government and urban protest in Colombia. Public Administration and Development, vol.8: 421-436. ---- (1989) Decentralisation and the need for political and critical analysis. Health Policy and Planning, 4(2): 168-171 ------ (1989) The rise and fall of the national 'decentralised agencies in Colombia. Public Administration and Development, vol. 9: 129-146. Conyers, D. (1983) Decentralisation: the latest fashion in development administration? Public Administration and Development, vol 3: 97-109. ---------- (1981) Decentralisation for regional development: a comparative study of Tanzania, Zambia, and Papua New Guinea. Public Administration and Development, vol. 1: 197-20 --------- (1986) Future directions in development studies: the case of decentralisation. World Development, vol. 14: 593-603. Cox, S. and Anise, S. (1982) Community Participation in rural water supply. Grassroots Development, 6:1: 3-7. Dave, P. (1991) Community and self-financing in voluntary health programmes in India. Health Policy and Planning, 6(1): 20-31. Davies, L. (1988) Contradictions of control: lessons from exploring teachers' work in Botswana. International Journal of Educational Development, vol. 8; no. 4: 293-303. Dillinger, W. (1989) External sources of local government finance, in EDI (1989). Douglas, D. (1990) In the vessel's wake - the UN Water Decade and its legacy. Grassroots Development, 14/2: 3-11.

EDI Policy Seminar Report (1989) Strengthening local governments in sub-Saharan Africa; no 21; World Bank: Washington DC. Elhussein, A.M. (1988) The revival of 'native administration' in the Sudan: a pragmatic view. Public Administration and Development, vol.9: 437-446. Garret, R.(1990) The introduction of school-based curriculum development in a centralised education system: a possible tactic. International Journal of Educational Development, vol. 10; no. 4: 303-309. Gilbert, A. and Ward, P. (1985) Housing, the state and the poor. Cambridge University Press: Cambridge, U.K. Gish, O., Malik, R. and Sudharto, P. (1988) Who gets what? Utilisation of health services in Indonesia. International Journal of Health Planning and Management, vol. 3: 185-196. Godhinho, J. (1990) 'Tipping the balance towards primary health care': managing change at the local level. International Journal of Health Planning and Management, vol. 5: 4152. Gonzalez-Block, N et al. (1989) Health services decentralisation in Mexico: formulation, implementation and results of policy. Health Policy and Planning, 4: 301-15. Goode, R. (1990) Obstacles to Tax Reform in Developing Countries, in Bird and Oldman, (ibid.) Harbeson, J. (1990) Centralisation and development in Eastern Africa, in Wunsch and Olowu (op. cit) Harriss, B. (1988) Government revenue and expenditure in an agrarian economy: North Arcot District in South India. Public Administration and Development. Hasan, A. (1988) Low cost sanitation for a squatter community. World Health Forum, vol. 9: 361-364. Havassy, H. and Yany, U. (1989) Bridging between local needs and centralised services: co-option or co-production? Community Development Journal, vol. 25 no. 3: 213 Heaver, R. (1984) Adapting the training and visit system for family planning, health, and nutrition programmes. World Bank Staff Working Papers, no. 662. Helm, D. and Smith, S. (1986) The assessment: decentralisation and the economics of local government. Oxford Review of Economic Policy, vol. 3; no. 2. Hinchcliffe, K. (1989) Federal financing of education: issues and evidence. Comparative Education Review, vol. 33; no. 4: 437-449.

Hopper, W. (1989) The response from the grassroots: self reliance in Zambian education. IDS Bulletin, 20(1): 17-23. Hughes, A. (1985) Alternative forms and levels of popular participation: a general survey, in Lisk (1985) Huque, A. (1986) The illusion of decentralisation: local administration in Bangladesh. International Review of Administrative Sciences, vol. 52: 79-95. Iyun, F. (1988) Inequalities in health care in Ondo State, Nigeria. Health Policy and Planning, 3(2): 159-163. Jack, W. (forthcoming) Power sharing and pollution control: co-ordinating policies among levels of government. World Bank PPR: Washington DC. Jimenez, E. (1988) Pricing policy in the social sectors. Johns Hopkins University Press: Baltimore U.S.A. Khan, M. (1987) Paradoxes of Decentralisation in Bangladesh. Development Policy Review: 407-412 King, D. (1988) Civil service policies in Indonesia: an obstacle to decentralisation? Public Administration and Development, vol. 8. Kowli, S., Bhalerao, V., Jagtap, A. and Shrivastaw,R. (1990) Community participation boosts immunisation coverage. World Health Forum, vol. 11: 169-172. Laleman, G. and Annys, S. (1990) Understanding community participation: a health programme in the Philippines. Health Policy and Planning: 251-256. Lauglo, J. et al (eds.) (1985) The control of education: international perspectives on the centralisation/decentralisation debate. London: Heinemann. Leung, J. (1990) The community-based welfare system in China. Community Development Journal, vol. 25 no. 3. Lewis, J. et al (1988) Strengthening the poor: what have we learned? Overseas Development Council: Washington DC. Lisk, F. (ed.) (1985) Popular participation in planning for basic needs. ILO: Gower, Aldershot. MacAndrews, C. (ed.) (1986) Central government and local development in Indonesia. Oxford University Press.

Mackintosh, M. and Wuyts, M. (1988) Accumulation, social services and socialist transition in the Third World: reflections on the Mozambican experience. Journal of Development Studies, 24(4) July: 136-179. Madhusudhan, R. and Berkhead, J. (1987) Expenditure trends in Selected Industrialised Countries. Public Budgeting and Finance, Winter 1987: 49-56. Maitra, T. (1982) Expansion of employment through local resource mobilisation. ILO: Asian Employment Programme: Bangkok. Majeres, J. (1985) Popular participation in decision-making with reference to developing planning: an institutional approach, in Lisk (1985). Manglesorf, K. (1988) Administrative decentralisation and development: some conflicting evidence from Ecquador. International Review of Administrative Sciences, vol. 54: 67-88 Manglesdorf, K., Luna, J. and Smith, H. (1988) Primary health care and public policy. World Health Forum, vol. 9: 509-513. Marlow, P. (1988) Fiscal decentralisation and government size. Public Choice, 56: 259269. Maro, P. (1990) The Imoact of Decentralisation on Spatial Equity and Rural Development in Tanzania. World Development, vol. 18, no. 5: 673-93. Mawhood, P. (ed.) (1983) Local government in the Third World. The experience of tropical Africa. John Wiley and Sons: Chichester. Mayer, J. (ed.) (1988) Bringing jobs to people. ILO: Geneva. McGinn, N. and Street, S. (1986) Educational decentralisation: weak state or strong state? Comparative Education Review, 30: 471-90. Miller, B. (1982) The Role of Women in the Public Domain in Developing Countries. International Supplement to the Women's Studies Quarterly. Mutiza-Mangiza, N. (1990) Decentralisation and district development planning in Zimbabwe. Public Administration and Development, vol. 10: 423-35. --- (1990) Lessons from Tanzania's experience of rural local government reform. International Journal of Public Sector Management, vol. 3; no.3: 23. Mwira, K. (1990) Kenya's Harambee secondary school movement: the contradictions of public policy. Comparative Education Review, vol. 34 no. 3: 350-368.

Narayan-Parker, D. (1988) Low-cost water and sanitation: tasks for all the people. World Health Forum, vol.9: 356-360. Nellis, J. (1983) Tutorial decentralisation in Morocco. Journal of Modern African Studies, 21: 3: 483-508. Nelson, M. (1986) An empirical analysis of state and local tax structure in the context of the Leviathan model. Public Choice, 49: 283-294. Newell, K. (1989) The way ahead for district health systems. World Health Forum, vol.10: 80. Nooi, P.S. (1987) Local Self-government - a comparative study of the countries of Africa and Southeast Asia. Planning and Administration, vol. 14(1)) Norris, N (1983) Local government and decentralisation in the Sudan. Public Administration and Development, vol. 3: 209-22. Nwankwo, G. (1984) Management problems of the proliferation of local government in Nigeria. Public Administration and Development, vol. 4: 63-76. Nyamekye, K. (1983) Community government in the North Solomons province. YaglAmbu, vol. 10 no.4. Oates, W. (1985) Searching for Leviathan: an empirical study. American Economic Review, vol. 75; no. 4: 748-757. Ojo, K. (1990) The crisis in the distribution of health personnel in Nigeria. Health Policy and Planning, 5(1): 60-66. Olowu, D. (1982) Local government innovation in Nigeria and Brazil: a comparative discussion of innovational transfers and intergovernmental relations. Public Administration and Development, vol. 2: 345-357. Olowu, D. (1987) African local governments since independence: which way forward? Planning and Administration Olowu, D. (1987b) Nigeria: federal and state transfers to local governments, 1970-1987. World Bank Working Paper Olowu, D. (1989) Local institutes and development. CJAS / RCEA 23:2: 201-231. Opuku-Afriyle, Y. (1985) Dilemmas of citizen participation and decentralisation. Greenhill Journal of Administration, 5: 16-24.

Organisation for Economic Co-operation and Development (1986) Rural Public Management. OECD: Paris. Page, D. (1984) Housing groups in Argentina. Grassroots Development; 8:1: 39-42. Parry, G. (1972) Participation in politics. Manchester University Press: Manchester, U.K. Pasha, H and McGarry, M. (ed.) (1989) Rural water supply and sanitation in Pakistan lessons from experience. IBRD: Washington DC. Paul, S. (1988) Governments and grassroots organisations: from co-existence to collaboration, in Lewis (op cit.) ---- (1989) Community participation in development projects. World Bank Discussion Paper, no.6: Washington D.C. Picard, L. and Zariski, R. (ed.) (1987) Subnational politics in the 1980s. Organisation, reorganisation and economic development. Praeger Publishers: N.Y, U.S.A. Pokawin, P. (1983) Community governments in Manus: people's participation in government. Yaglu-Ambu, vol. 10 no. 4. Premdas, R. (1988) Decentralisation, development and secession: the case of Papua New Guinea. Marga, vol. 9 no. 3. Prud'homme, R. (1989) Main issues in decentralisation, in EDI (1989) Rakodi, C. (1988) The local state and urban local government in Zambia. Public Administration and Development, vol.8: 27-46. Ralston, D. (et al) (1983) Voluntary Efforts in Decentralised Management. Berkeley: California Robinson, J. (1986) Decentralisation, money and power: the case of people-run schools in China. Comparative Education Review, vol. 30. no. 1: 73-88. Rondinelli, D., Nellis, J., and Cheema (1983) Decentralisation in Developing Countries a review of recent experience. World Bank Staff Working Papers; no. 581: Washington DC. Rondinelli, D. and Wilson, P. (1987) Linking decentralisation and regional development planning: the IRD project in Peru. American Planning Association, Summer issue: 348357.

Rondinelli, D., McCullogh, J. and Johnson, R. (1989) Analysing decentralisation policies in developing countries: a political economy framework. Development and Change, vol.20: 57-87. Rondinelli, D. (1990) Financing the decentralisation of urban services in developing countries: administrative requirements for fiscal improvements. Studies in Comparative International Development, vol. 25: 43-59. ----- (1990) Decentralisation, territorial power and the state: a critical response. Development and Change, vol.21: 491-500. Rosen, S. (1985) Recentralisation, decentralisation and rationalisation: Deng Xiaping's bifurcated educational policy. Modern China, vol.11: 301-46. Ruiz-Gallegos, M. (1990) Decentralisation and co-ordination: two techniques of serving the general interest. International Review of Administrative Sciences, vol. 56: 267-284. Ruland, J. and Sajo, T. (1988) Local government and development in a regional city: the case of Iloilo City, Philippines. Public Administration and Development, vol. 8: 261-287. Samoff, J. (1990) Decentralisation: the politics of interventionism. Development and Change, vol. 21: 513-530. ---- (1990) The politics of privatisation in Tanzania. International Journal of Educational Development, vol. 10; no.1: 1-15. Sanwal, M. (1987) Implementation of decentralisation: a case study of district planning in Agra District, Uttar Pradesh, India. Public Administration and Development, vol. 7: 38397. Schaffer, W. (1982) Institution building for development administration for rural areas in Africa Schmidt, W. (1989) Political variables and governmental decentralisation in Peru, 194888. Journal of Inter-American Studies and World Affairs. Schroeder, L. (ed.) (1989) Financing governmental decentralisation - the case of Bangladesh. Westview Press: Colarado, U.S.A. Silanda, E. and Tuinjnman, A. (1989) Regional Variations in the financing of primary education in Zambia. International Journal of Educational Development, vol. 9; no. 1: 518. Skinner, R. et al (ed.) (1987) Shelter upgrading for the urban poor. Island Publishing for HABITAT and BIE: Manila, Philippines.

Slater, D. (1989) Territorial Power and the Peripheral State: The Issue of Decentralisation. Development and Change, vol. 20 no. 3: 501-33. Slater, D. and Watson J. (1989) Democratic decentralisation or political consolidation: the case of local government reform in Karnakata. Public Administration and Development, vol. 9: 147-57. Smith, B. (1982) the revenue position of local government in Nigeria. Public Administration and Development, vol. 2: 1-14. Smith, B. (1986) Spatial ambiguities: decentralisation within the state. Public Administration and Development, vol. 6: 455-65. Sorock, M. (1984) Self-help housing from Colombia: a case study from Cartago. Grassroots Development, vol. 8, 1: 44-51. Standish, B. (1983) Power to the people? Decentralisation in Papua New Guinea. Public Administration and Development, vol. 3: 223-238. Stren, R. (1989) Institutional arrangements, in E.D.I. (1989) Struyk, R. (1988) Assessing housing needs and policy alternatives in developing countries. Urban Institute Report, 88-4. Tarimo, E. and Fowkes, F.G. (1989) Strengthening the backbone of primary health care. World Health Forum, vol. 10: 74. Thomas, P. (1989) Local level planning in Bangladesh. Paper commissioned by UNDP. Thomason, J. (1988) Disbursement, decentralisation and development: lessons from the first rural health services project in Papua New Guinea. Public Administration and Development, vol. 8: 391-99. Tilak, J. (1989) Centre-state relations in financing education in India. Comparative Education Review, vol. 33; no. 4: 450-479. Tordoff, W. (1988) Local Administration in Botswana. Public Administration and Development, vol. 8: 183-202. Tsai, H. (1990) Population decentralisation policies: the experience of Taiwan. Industry of Free China: 1-12. Tumwine, J. (1989) Community participation as myth or reality: a personal experience from Zimbabwe. Health Policy and Planning, 4(2): 157-161.

Unger, J., Mbaye, A. and Diao, M. (1990) From Bamako to Kolda: a case study of medicines and the financing of district health services. Health Policy and Planning, 5(4): 367-377. UNICEF (1990) The state of the World's children. Oxford University Press: Oxford, U.K. United Nations (1989) Housing and economic adjustment. Taylor and Francis: New York. Uphoff, N. (1988) Assisted self-reliance: working with, rather than for the poor, in Lewis (1989) U Tin U et al (1988) "We want water not gold". World Health Forum, vol. 9: 519-525. de Valk, P. and Wekwete, K. (ed.) Decentralisation for participatory planning? Comparing the experiences of Zimbabwe and other Anglophone countries in eastern and southern Africa. Aldershot: England, U.K. Vengroff, R and Johnston, A. (1987) Decentralisation and the implementation of rural development in Senegal: the role of rural councils. Public Administration and Development, vol. 7: 273-88. Wallis, M. (1990) District planning and local government in Kenya. Public Administration and Development, vol. 10: 437-452 Wasylenko, M. (1987) Fiscal decentralisation and economic Development. Public Budgeting and Finance: 57-69. Wellings, P. (1988) International competition and industrial decentralisation in South Africa: a comment. World Development, vol. 16; no. 12: 1547-1549. Werlin, H. (1990) Decentralisation and culture: the case of Monrovia, Liberia. Public Administration and Development, vol. 10: 251-61. Wilson-Pepper, C. R. (1982) The place of local social realities in developmental work. Public Administration and Development, vol. 2: 195-205. Winkler, H. (1989) Decentralisation in education: an economic perspective. World Bank PPR Working Paper. Wirtshafter, R. and Shih, E. (1990) Decentralisation of China's Electricity Sector: Is Small Beautiful? World Development, vol 18, no. 4: 505-12. World Bank (1990) Primary Education. World Bank Policy Paper: Washington, DC. World Development Report (1983) World Bank: Washington DC.

World Health Organisation (1991) Community involvement in health development: challenging health services. WHO Technical Report Series. Wunsch, J. and Olowu, D. (ed.) (1990) The failure of the centralised state. Westview Press, Colarado, U.S.A. Wunsch, J. (1990) Beyond the failure of the centralised state: toward self-governance and an alternative institutional paradigm. in Wunsch and Olowu (1990). Wunsch, J. (1991) Sustaining Third World infrastructure investments: decentralisation and alternative strategies. Public Administration end Development, vol. 11: 5-23. Yoder, R. and Eby, S. (1985) Participation, job satisfaction and decentralisation: the case of Swaziland. Public Administration and Development. Yudelman, S. (1986) After Nairobi: a retrospective of women's development organisations in Latin America. Grassroots Development, vol. 10: 1: 2029.

You might also like